DEV Community: baiwei

Creating a Birthday Surprise Page in One Sitting

baiwei — Thu, 29 Oct 2026 17:06:22 +0000

Planning a meaningful birthday surprise does not have to take days of complicated effort. When you want to celebrate someone who matters to you, gathering memories and kind thoughts into a single digital space creates an experience that feels personal, deliberate, and lasting. Many people find themselves searching through disconnected photo collections, text conversations, and audio tracks, wondering how to bring these elements together without turning the project into an overwhelming chore. By sitting down with an orderly sequence in mind, you can assemble a complete presentation in a single quiet afternoon or evening. Using a dedicated birthday surprise website maker like Birthday Website Maker makes it straightforward to bring everything together smoothly from start to finish.

The most effective starting point is the heartfelt letter. Before choosing any visual media or audio, spend a few uninterrupted moments putting your thoughts onto the page. Focus on genuine memories, specific qualities you admire in the recipient, and your warmest hopes for their upcoming year. Writing the words first grounds the entire project, establishing an authentic emotional tone that guides every choice that follows. When the text is sincere and complete, the rest of the page simply serves to support and elevate what you have written, rather than competing with it for attention.

Once your message is ready, the next natural step is selecting the photos. Choose images that reflect meaningful moments rather than trying to document every single event. Candid snapshots, shared trips, and everyday smiles often carry more warmth than posed portraits. After the visual story is in place, introducing music sets the atmosphere. Sound adds an emotional layer, turning a quiet reading experience into something richer and more immersive. Visiting Birthday Website Maker – Create a Birthday Surprise Website allows you to bring these three core elements—your photos, background music, and heartfelt text—together into a single, cohesive presentation.

With the content assembled, reviewing the result is essential. You can take advantage of a free preview to see exactly how your creation looks and flows. Walk through the page from top to bottom just as your recipient will experience it. Check the pacing between the photos, ensure the music begins smoothly, and confirm that your letter reads comfortably from beginning to end. Having the ability to preview free gives you peace of mind, allowing you to catch minor adjustments before you unlock the final version. When you decide to Create a Birthday Surprise Website, this preview step ensures that nothing feels rushed or unfinished.

The final step is unlocking the completed page and sharing it. Instead of coordinating multiple files or sending separate media attachments across different messaging apps, you simply share one link. The recipient opens that single URL and immediately steps into a focused, personal celebration crafted specifically for them. It requires no technical knowledge on either side, providing a direct, memorable surprise that can be revisited whenever they wish to remember the occasion. Sitting down to create a personalized birthday website with photos, music, and a heartfelt letter turns what could be a scattered effort into a calm, rewarding creative ritual.

Developer Field Notes on Preparing a Static Document for Rapid Text Revision

baiwei — Thu, 29 Oct 2026 02:07:30 +0000

Every developer encounters a moment where an incoming specification or project memo arrives locked inside a PDF. When the immediate goal is to draft changes, rewrite paragraphs, or reorganize explanations, working across a rigid visual layout creates needless friction. Copying fragmented snippets out of a visual viewer often results in broken sentence flows, unwanted hyphenations, and awkward line breaks that interrupt creative momentum. Before any substantive editing can begin, the text needs to exist in an environment where sentences can move freely. Converting that initial document into a lightweight markup format bridges the gap between a frozen page and an active text workspace.

To streamline this initial preparation step, I rely on a browser window opened to PDF to MD Converter — Free Online Tool. The utility is built entirely around an uncomplicated premise: taking a text PDF and producing clean, editable plain text without unnecessary hurdles. There is no account required to begin, which removes any onboarding delay when I only have a few minutes between tasks. Because it functions as a free in-browser pdf to md tool, the workflow feels natural alongside my daily development tools.

Security and data boundaries are always a primary consideration during development work. Many project summaries and technical briefs contain internal discussions or early notes that should not travel anywhere outside the local environment. With this utility, files stay on your device from start to finish. The browser handles the reading process directly within the client tab. Knowing that nothing leaves the personal workstation provides confidence when processing sensitive documentation. The entire utility, known as PDF to MD, focuses squarely on handling the file right where it is.

The mechanics of the transition are direct. All that is needed is to drop a text PDF into the designated area. Within moments, the interface displays an immediate preview Markdown section. Seeing the text rendered cleanly in the PDF to MD Converter allows for a quick visual sanity check. I can glance through the sentences, verify that the passage flows continuously, and confirm that the paragraphs are intact before moving forward. There is no complex configuration or convoluted settings panel to navigate. The screen simply presents what has been converted, giving you full control over the result.

Once the preview looks right, the next step depends on where the text is headed. The application provides two straightforward options: you can copy or download the generated text. If I am pasting a specific paragraph into an ongoing technical scratchpad, copying from the preview is instantaneous. If the document represents an entire article or documentation draft that requires long-form revision, downloading the file directly to my project directory is the better path. Relying on pdf to md turns what used to be a frustrating copy-paste chore into a seamless handover.

Having clean Markdown ready on your local machine completely shifts the focus back to editing the text. Sentences can be rearranged, phrases can be rewritten, and sections can be expanded in your favorite plain text editor without wrestling with print artifacts. Taking a few seconds to run an incoming text PDF through a dedicated browser utility before touching the prose saves significant clean-up effort later. It keeps the workspace tidy, maintains strict local data privacy, and lets you concentrate purely on refining your words.

Developer Field Note: Pipeline Inputs and Texture Failure Checks

baiwei — Tue, 27 Oct 2026 21:40:01 +0000

When routing raw diffuse textures into a 3D asset pipeline, local texture generation prevents workflow friction and keeps project assets secure. Loading a JPG, PNG, or WebP image into a Free normal map generator with live 3D preview executes entirely within your browser client. Files stay on your local hardware without account signups, external server uploads, or image watermarks.

Beyond tangent normal maps, an Image to Normal Map Converter generates a complete material set: albedo, roughness, ambient occlusion, metallic, bump, curvature, specular, and high-precision 16-bit PNG height and displacement maps. Fine-tuning strength values while inspecting geometry across a shader ball, cube, or flat plane reveals issues before baking meshes. In practice, successful engine handoff hinges on two specific failure checks: the Y sign convention and color space rules.

The primary failure point is the green channel vector sign. Normal maps follow either OpenGL (+Y green) or DirectX (-Y green) coordinate standards. Blender, Unity, and Godot expect OpenGL conventions, while Unreal Engine strictly requires DirectX. If lighting directions look inverted and surface bumps register as sunken dents on your mesh, toggle invert height or swap the normal green axis to match the renderer.

The secondary failure point is texture color space handling upon import. Engine ZIP downloads organize textures cleanly for Unity, Unreal, Godot, Blender, or Generic workflows, but material samplers still require precise verification. Unity needs albedo in sRGB, normal maps explicitly assigned as normal type, and MetallicSmoothness packed with metallic in R and smoothness in A with sRGB disabled. Unreal demands sRGB enabled on base color, normal compression on normals, and sRGB unchecked for packed ORM masks. In Blender, keep base color in sRGB, but switch normal, roughness, metallic, and displacement texture nodes to Non-Color data. Refer to Import PBR Maps into Unity, Unreal Engine and Blender to keep shading mathematically accurate across every target environment.

Solid, Broken, Changing: A Record Spec for the I Ching

baiwei — Tue, 20 Oct 2026 13:57:34 +0000

Eliot Weinberger wrote the essay this account follows. It ran in The New York Review of Books. The issue date is 2016-02-25. He asked how one volume can be pressed into service as an explanation of everything and still withhold a single settled sense. The names, the dates, and the clashing English versions below come from that essay. The reading order does not. Start with the marks you would actually store. The sermons people have hung on those marks can wait until the fields are named.

A line here is not a sentence. It is one of two glyphs. The solid glyph is yang. The broken glyph is yin. Six ordered slots, two glyphs, and nothing else in the figure itself: two to the sixth power is sixty-four. That count is the whole address space. Commentary never enlarged it. What commentary did, for centuries, was argue about the strings written beside the address.

Two glyphs, eight letters, sixty-four addresses

The smaller unit is the trigram. Three slots, still solid or broken, give eight trigrams and then stop. The set is an alphabet, not a list you can append to. Tradition hands each letter a nature-name. Put heaven and earth at the poles. The remaining letters are lake, thunder, and fire, and also water, mountain, and wind. Same eight. Not the same sequence you will see in older tellings, and the sequence of the names is not the encoding. The encoding is the stack.

Fu Xi is the myth attached to that alphabet. Pictures give him a human face on a serpent's body, or a dragon's body. He is said to have watched star-patterns, cloud-motion, breaks in rock, and marks on birds and beasts until the variety collapsed into eight trigrams, read as yin and yang. Out of the same letters he is said to have taught farming, kingship, marriage, writing, and how to find a route. You can refuse the hero and still count the code.

Stack one trigram on another and the figure has six lines. Eight times eight lands on the same sixty-four hexagrams you get by counting six binary places. Direction is data. If the two letters differ, flipping the stack addresses another member of the set. Hexagram 52 is a poor test of that rule, because mountain set on mountain still looks like itself after a vertical flip.

Legend dates the stacking to about 1050 BCE and gives it to Emperor Wen, founder of the Zhou house. He is said to have numbered all sixty-four combinations, arranged them, named them, and attached a short oracle to each. Later readers call those oracles Judgments. The notes on single lines, called simply the Lines, are credited to his son, a poet, the Duke of Zhou. Even if the calendar is doubted, the genealogy keeps three layers apart: the figure, the oracle for the whole figure, and the text for one line.

Five centuries on, Confucius was credited with ethical commentaries on every hexagram, the Ten Wings. "Wing" here is a builder's word, a side hall, not a feather. Wings attach to the six-line figure. They are not the figure. It is nearly sure that Confucius took no part in composing the book. A saying still circulates that, given another century of life, he would have spent half of it on this text. That admits difficulty. It does not fill an author column.

The marks are older than that family tree and older than the glossary. Shang rule is placed near 1600 BCE, and the practice may reach further back. Diviners heated tortoise shells, or ox shoulder bones, and read the cracks. Oracle bones have come out of the ground in the hundreds of thousands. Plenty carry a finished hexagram, or a number assigned to one. Where the figures came from, and how a crack became a reading, is unknown. Encoding and the book that explains it did not ship together.

What an entry must hand over, and what a cast adds

Before the Western arguments, write the columns down. Almost every quarrel in the rest of this piece is a quarrel about which column a sentence belongs in.

Once the Han court had fixed a standard, a hexagram entry carried at least the following.

A key: a number plus a name. The number is stable only inside one ordering of the sixty-four. The name is a word, and the words in this book move.
A figure: two trigrams, lower and upper, drawn as six lines. When a cast produces the figure, the earliest result occupies the lowest slot. Call that line 1 if your notes count from one. The last result is line 6, at the top. Position is meaningful. A remark about the calves is not the same remark moved onto the jaws for the sake of rhythm.
A Judgment: a short oracle for the whole figure, often opaque.
An Image: a reading of the picture made by the two trigrams. Mountain set over mountain is already a picture, before anyone extracts a moral.
Six line texts, one per slot, commonly stranger than the Judgment.
Commentary that may be absent from the file you opened. The Ten Wings are the classical comment layer. A modern translator may print them, digest them, swap in a later scholar, or leave them out.

A cast is a second object. The printed book shows one hexagram at a time. A consultation adds columns the page does not store in advance.

The question, supplied by the reader. Your question is not in the text.
A procedure that selects the lines. The classical pair are three coins, or fifty yarrow stalks.
Lines that change. They hold the sharpest counsel, and they invert. After the inversion you are looking at a second address, the relating hexagram, which the tradition reads as the way the situation is tending.

That is why "six bits" describes the drawn glyph and does not describe a consultation. The glyph is binary, six times. The consultation can also flag a line as moving. Primary figure, set of moving lines, relating figure: that is the result record. A translation of the still book specifies the book. A reading tool has to specify the cast. Mix the two, and a confident paragraph will borrow a field from the wrong layer and present it as the old line.

cast:
question: reader state, outside the book
procedure: three coins, or fifty yarrow stalks
changing: positions that invert; the list may be empty
primary: address of the six marks as cast
relating: address after those positions flip




There is no neutral English for this layout. James Legge, Richard Wilhelm, John Minford, David Hinton, Richard John Lynn, Richard J. Smith, and Arthur Waley are not offering seven styles of one string. They decide, often without saying so, whether the Image may be folded into the Judgment, whether a Han commentator may speak as "the text," whether a Latin tag may sit inside a Zhou oracle, and whether a pictographic etymology may replace the received gloss. Those are specification choices. The history that follows is a history of people making them.

One collision shows why no default English noun was available. Inside the *I Ching*, a single word can name a captive taken in war and can also name sincerity. The captive reading leans toward violence and political fate. The sincerity reading leans toward an inner quality. Both sit in the Chinese tradition. They do not compile to one program, so keeping a single English noun is a product choice and should be labeled as one.

The coin procedure is the cast in its cheap form. Labor is the only cost worth recording. Shells needed a specialist to perform the rite and to read it. Anyone could throw three coins, and anyone could count fifty yarrow stalks. Faces carry numbers. The three coins go up together, and the total of a throw picks a solid line or a broken one. Six throws, six lines: six tosses make the hexagram. Stalk counting is the slow route to that same output. The chances on each throw, and the rule that the first result sits in the lowest slot, are laid out in [steps for the three-coin method](https://ichingonline.pro/blog/three-coin-method). Neither tool prints a meaning. Each tool emits lines.

## Cracks, an oral book, a canon

At some point in the Zhou centuries, the current guess being about 800 BCE, the sixty-four hexagrams received names and a written text condensed out of oral telling. The book was the *Zhou Yi*, the Changes of Zhou. A plastron or a bone stayed an expert's instrument. Coins and yarrow stalks could be used by whoever would perform them, and both are still in use. The output type did not change when the tool did. Only the crack has lost its interpreter.

By the third century BCE, with Confucianism rising, the Ten Wings had been fastened on. The *Zhou Yi* was no longer only a diviner's manual. It had become philosophy and ethics. Han emperor Wu, in 136 BCE, set this volume above the other four Confucian classics and locked one version in place. Rivals remained, some with a different order of figures. Order is part of the spec: the same sixty-four states in another sequence are another index, and number 52 opens only the sequence you standardized. The locked book is the *I Ching*, the Classic of Change, also called the Book of Change. From then on an entry keeps its shape: name, number, a hard Judgment, an often poetic Image taken from the two trigrams, and a dark statement for each of the six lines.

For about two thousand years that small shape was treated as the necessary guide to everything. The philosophy around it said things are tied together and never sit still. The book did not describe the universe from outside. It was taken for the universe's closest miniature. One Sinologist's phrase was "underpinnings of reality." The sixty-four hexagrams hardened into bins other fields were expected to use. Judgments were handled as seeds. Wings and later commentaries grew them into counsel for a ruler who wanted order, and for everyone else who wanted a proper life. The same object was used to meditate, and, because it was said to belong to how the world works, to ask what was coming, favorable or not.

Keep the code and the authority apart. The code is small: eight trigrams, six lines, sixty-four hexagrams, two glyphs. The authority is a repeated decision to treat that code as a total index. Once an observation is coarsened until it fits, a closed set of labels can swallow it. That is how literary criticism and picture criticism could use the book, how mapmakers could, how medicine could, and how several of the sciences could, either as a sorting frame or as a closing proof. Fitting is not the same thing as being precise.

Schools never agreed. Confucians fought among themselves, Taoists did the same, Buddhists added further lines, and Christians arrived later with more. In China and across East Asia no book was consulted more often, or trusted on a wider spread of questions. Western readers have had it for over three centuries. From the 1950s on, it is fair to call it the Chinese book the wider world is most likely to recognize. Weinberger's picture for the center is a void, or a cloud whose outline will not stay put, because the words that do the real work never froze into one sense.

## Overwriting the value column

Jesuit missionaries in China had the book by the late seventeenth century and read it as Christian truth with a universal claim. They kept the indices and replaced the values. Slot 1 became God. Slot 2 became Jesus, taken as a second Adam. Slot 3 became the Trinity. Slot 8 became Noah's household. The series went on in that spirit. Anyone who needs the book to confirm a system already held will run a version of the same cast.

Leibniz, with real excitement, recognized his own binary scheme in the solid marks and the broken ones. His move was narrower than the Jesuit move. He was matching glyphs to a notation he already had, not matching hexagram numbers to scripture. The recognition is real. Two marks, arranged in stacks, are a binary notation. It does not follow that the book is a machine, a program, or a theory of computation. Later centuries wanted that stronger claim. Leibniz does not file it for them. He shows that a mathematician can see the encoding and stop before the oracle.

Hegel, who did not think Confucius repaid the work of translation, brushed the book off as "superficial." What he said he failed to find was a sensuous picture, even in a single case, of universal powers whether natural or spiritual. He wanted universals offered in sensory form. A reader who wants definitions has a different complaint. The book is almost nothing but a formal grid, and the prose on the grid will not behave like a definition. Hegel's irritation and a specifier's irritation are not one irritation. Both come from asking the file for a genre it does not have.

The earliest English translation is by Canon Thomas McClatchie, an Anglican cleric posted in Hong Kong. Weinberger sets him beside Reverend Casaubon. *Middlemarch* had been out four years when, in 1876, McClatchie announced that every mythology now had its key and that he held it. On his telling the *I Ching* reached China with a son of Noah, and the book was an obscene hymn to a hermaphroditic monad, which he matched with Chaldean Baal and with Hindu Shiva. Every sign collapses into the one sign the reader brought. The binary grid sits underneath, idle.

James Legge, a Hong Kong missionary like McClatchie, called that performance "delirious." He disliked China in a broad way, and he remains the earliest English translator who matters for the Chinese classics as a body. The job took twenty years, broken by other work. A wreck in the Red Sea took the manuscript down. In 1882 he issued an English *I Ching* later readers could lean on, unlike the 1876 performance. He was first to call the stacks hexagrams, taking the English word from the star that has six points. Readers now hear a star in a name that, on the page, is six piled lines. The installed word stays.

Legge was professionally horrified by what he took for idolatry and superstition, and he still found the book pulling him in, a fascination he described as gradual and strong. Out of that pull he built a theory of translation. Chinese characters, he argued, were "symbols of ideas" rather than "representations of words." Set in a composition, they showed the writer's thought, not the sentence the writer would have spoken. The translator was supposed to come "*en rapport*" with an author: minds meeting past the written line. This book has no author for that meeting to reach. Legge's version is full of clarifications anyway, dropped into parentheses inside a rendering that is otherwise literal.

Herbert Giles, the next English translator of consequence after Legge, took the *I Ching* for "apparent gibberish." Educated Chinese, he said, would freely grant that the pages looked like nonsense, and would still insist that serious lessons waited for anyone sharp enough to draw them out. Unreadable and mandatory at once: that is why another commentary always looks like a reasonable reply to failure.

Waley's 1933 study stopped short of a complete translation. He called what he had a pile of omens under "peasant interpretation," with specific divinations grafted on later. Tell his example backward, so the migration stays visible. A diviner's rule: a red morning sky is inauspicious, so do not cross by boat. Behind the rule is an older weather saying, shepherds treating a red dawn as a warning. The sky is unchanged. The sentence has gained a prohibition. An omen is a sign. A divination binds that sign to an act. Waley sorted the signs into three bins: feelings and movements nobody chooses, twitches and stumbles and belches among them; signs from plants, animals, and birds; and signs from thunder, stars, and rain. He found all three, and he refused metaphysics on purpose.

Joseph Needham gave the *I Ching* many irritated pages in *Science and Civilisation in China*. He called it a "pseudo-science" that harmed real Chinese inquiry for centuries, because exact notes on nature and on physical things were forced into the hexagrams' vague boxes. His name for the boxes was a "cosmic filing-system." The mechanism is usable without his whole verdict. A closed, socially mandatory set with soft edges will take in an observation it cannot distinguish, and hand back a label that feels like an explanation. He was judging a system that had been allowed to classify the physical world, not reviewing a book of poems.

## A merged build that became the public text

Richard Wilhelm's German of 1924, and especially Cary F. Baynes's English of 1950, carried the book out of specialist hands and made it famous. Unlike Legge, this missionary was a partisan of the Wisdom of the East and ranked China first inside it. Western life, as he saw it, had been mechanized and rationalized without letup. It needed what he took for the East's hold on a natural depth of soul. He wanted the sides joined in mutual completion, on shared human foundations, from a core inside the humane.

His rendering leans on Neo-Confucian comment from the late Song. Referents that were specifically Chinese get widened, in the name of a universal reader. The German printing carried crowds of notes claiming parallels with the Bible, with Kant, with Goethe, and with German Romantic writing. The English printing dropped them. The book is printed twice, once beside brief comments and once beside longer ones. Those comments fuse Chinese works with Wilhelm's own meditations and do not mark the joints. Needham filed the edition under a "Department of Utter Confusion," and added that Wilhelm seemed to be the one person who had the whole thing worked out. If you need to know which sentence is Zhou, which is Song, and which is Wilhelm, the missing seams are the defect.

Carl Jung supplied the introduction. Wilhelm took him for a mind in touch with Eastern findings and with the oldest Chinese wisdom he knew, and he matched Jung's *animus* and *anima* to *yin* and *yang*. One assertion now reads as painful: Jung treated it as curious that a people this gifted had never produced what he meant by science. He also set chance, renamed synchronicity, as the guiding idea of a sacred book. Firm believers answer that the *I Ching* does not run on chance at all. Coins still land either way. A margin that prints "synchronicity" has chosen Jung. A margin that refuses the word has chosen the believers. Neither label is a property of a solid line.

Through the 1950s and 1960s the Wilhelm/Baynes Bollingen volume was a sensation. Jorge Luis Borges, Octavio Paz, Charles Olson, and Allen Ginsberg were among those who turned its phrasing into poems. Fritjof Capra reached for the book inside *The Tao of Physics* when he wanted to talk about quantum mechanics. Terence McKenna thought the geometry lined up with the chemical waves that hallucinogens set off. Some readers took the two-mark notation as an early sketch of computing machinery. Philip K. Dick and Raymond Queneau each built a novel on the book. Jackson Mac Low and John Cage built strict procedures that let the book emit poems, or pieces of music.

The excitement is intelligible without the metaphysics. People who preferred a runnable procedure to a finished art object had a generator. People wary of authority met a religious shape with no exotic gods and no priests, a cycle with no required start or finish, and no author. It arrived from an East many Westerners were inventing as the counter-image of their own civilization. Wilhelm's label for the cargo was "the seasoned wisdom of thousands of years." You could divine alone. It behaved like counsel for readers who would have refused any book shelved as self-help, and it offered a personal future. Bob Dylan's praise was blunt: for him this was "the only thing that is amazingly true."

## Minford's bulk and Hinton's short page

The two recent translations Weinberger sets side by side barely resemble each other. Set together, they work as a complementary *yin* and *yang*. One is maximal and documentary. The other is short, and it aims at a kind of selflessness. Neither is a patch on Wilhelm. Each is a different product on the same six-line substrate.

John Minford is known first for the five-volume English of *The Story of the Stone*, also called *The Dream of the Red Chamber*, generally ranked the greatest Chinese novel. David Hawkes started that project. Minford carried it on. His *I Ching*, plainly many years of work, runs over 800 pages, much of the type small. One binding holds two full translations. The "Bronze Age Oracle" rebuilds the Zhou text from before the later Confucian commentaries. The "Book of Wisdom" is that text as later centuries unpacked it. Each slice has an exegesis digesting older comment, earlier translators, and his own links into poetry, art, ritual, history, philosophy, and myth.

What matters for a record is that the two layers stay apart, joined by hexagram number rather than melted into one paragraph. You can dispute the rebuilt early layer. You can at least see which layer you are reading. Wilhelm's amalgam does not allow that.

David Hinton is grouped with Arthur Waley and with Burton Watson: literary Sinologists who answer to contemporary English. Watson is older by about a generation, and across the last fifty years the pair have mattered most in America for classical Chinese poetry and philosophy. Hinton's *I Ching* goes the other way from Minford's bulk: two pages per hexagram, a few Ten Wings excerpts, and almost no further voice after a short introduction. Read it straight through, like recent poetry, while it remains a translation rather than an imitation. Open it anywhere, as people open E. M. Cioran, or Elias Canetti.

He holds to a Taoist reading, or a Ch'an (Zen) one, and he is not interested in Confucian ethics or politics. The English he wants seats you in nature's Tao, the course of things as the living world enacts it and as he thinks the book enacts it too. He treats the hard Judgment lines as kin to later Taoist and Ch'an writing, devices meant to tease the mind past ordinary assumptions and past the cage of identity. That aim reverses Wilhelm's Jungian hope of realizing a self. Hinton is after selflessness. He also assumes that early Chinese life, closer to the ground and to land that still held wild country, was less cut off from nature's Tao. The assumption is his premise, not a line in the Judgments.

He sometimes reads the oldest graphs as pictures. Ezra Pound had already tried it on the *Book of Songs*, an early collection of Chinese poems, issued by him as *The Confucian Odes*. For Hexagram 32, usually Endurance, Duration, or Perseverance, Hinton prints "Moondrift Constancy," because the graph shows a half-moon held between a stroke above and a stroke below. Observation becomes "Heron's-Eye Gaze": a heron and an eye are both in the character, and his warrant is that a wading bird misses very little. He does this rarely. Weinberger notes that Sinologists will be scandalized. The odd English labels the method. A familiar smooth word would have hidden it.

## Gen, number 52, as a conformance fixture

One hexagram makes the distance measurable. Number 52 is Gen. The figure is not the part under dispute. It is the Mountain trigram doubled, one mountain set on another. In the received drawing, Mountain is a solid line over two broken lines. Read from the lowest slot upward, the six lines of Hexagram 52 are broken, broken, solid, broken, broken, solid. That sequence is an address. It is not yet a doctrine. The address stays put across the translations below. The English verb does not.

Judgment, comment tradition, and the six line texts for this figure are gathered on the page for [Hexagram 52: Keeping Still](https://ichingonline.pro/hexagrams/52-gen-keeping-still). Keeping Still is the name Wilhelm made familiar, so familiar that many English readers think it is simply what the hexagram is called. It is one legal name among several. The exercise is to hold the figure still while the names move.

![Hexagram 52 drawn as mountain stacked on mountain, plus a motionless person whose back faces us in a vacant court.](https://i.imgur.com/vRkR39v.jpeg)

What follows is a diff, not a ranking.

In the wisdom layer Minford's name-string is "Mountain," tracking the doubled trigram instead of leaping to a mood. His English stays spare, near telegraphese, and he centers lines rather than setting them flush left. A centered line looks like an inscription. He also allows Latin tags from early Jesuit versions inside the entry. His defense is that they ease the approach to a text this old and this foreign, set a contemplative mood outside any one century, and hint at a loose tie between Chinese self-knowledge and Europe's long concern with gnosis.

In the "Book of Wisdom" his Judgment for Hexagram 52 reads:

> The back
> Is still
> As a Mountain;
> There is no body.
> He walks
> In the courtyard,
> Unseen.
> No Harm,
> Nullum malum.

He then writes at length, and well, on mountains as spiritual fact and as poetic image in China. The Latin is not in the Chinese line. *Nullum malum* repeats "No Harm" in an older European specification. If strings are extracted by machine, that token has to be marked as Minford's interface, or the index will think the oracle switched languages.

Hinton's name is "Stillness." He sets the Judgment as ordinary sentences:

> Stillness in your back. Expect nothing from your life. Wander the courtyard where you see no one. How could you ever go astray?

"Expect nothing from your life" and "How could you ever go astray?" are clauses of counsel Minford does not print. His lines name a back, a mountain, a missing body, a courtyard, an unseen walker, and an absence of harm. Hinton reads the scene as instruction about expectation. "Stillness" already interprets the doubled mountain. "Mountain" stays closer to a type name. They answer different questions.

The title Wilhelm prints is "Keeping Still, Mountain." His Judgment:

> KEEPING STILL. Keeping his back still
> So that he no longer feels his body.
> He goes into the courtyard
> And does not see his people.
> No blame.

Store the gloss under his name, not as the line. Real quiet, for him, is not a freeze: stop while the moment asks for stopping, and move when it asks for motion, so rest agrees with the time and there is light in a life. He takes the figure as both the end of movement and its restart. The back is named, he says, because nerves that govern motion run there. Halt them and the restless ego drops away, after which a person can face outward and act with larger laws. Spinal nerves are a modern picture, not a Zhou term.

Richard John Lynn's edition, from Columbia University Press, calls itself definitive and treats the English that came before it as a long run of error. Lynn's word for Gen is "Restraint." The Judgment:

> Restraint takes place with the back, so one does not obtain [sic] the other person. He goes into that one's courtyard but does not see him there. There is no blame.

The explanation follows the Han commentator Wang Bi. Picture two people turned back to back. They can stand near each other and still fail to see each other. Neither one restrains the other. Each restrains himself.

"Restraint" is social and ethical. "Stillness" is physical and contemplative. "Mountain" is the trigram's own name. "Keeping Still" joins a verb of quiet to the mountain. Four English types, and the sacrificial readings have not arrived. Lynn's "[sic]" is a small honesty: he leaves "does not obtain the other person" unidiomatic rather than sanding it flat. The strain is data.

The six line statements of Hexagram 52 walk a body: feet, then calves, then the waist, the trunk, and at the top the jaws. Lower lines speak of lower parts, even while the verb stays in dispute. Wilhelm, with no historical warrant, guesses that the wording might be coaching something like yoga. A similar map of attention is not the same institution. The guess is his, and if it belongs anywhere it belongs in a note.

Line 2 is the calf line. It is where the implementations can be checked against one another without a general theory.

Hinton's second line:

> Stillness fills your calves. Raise up succession, all that will follow you, or you'll never know contentment.

Minford, in the wisdom layer, gives:

> The calves are
> Still as a Mountain.
> Others
> Are not harnessed.
> The heart is heavy.

His note hears a possible healing in the stillness, then says the energy of other people will not be brought under control. Retreat is not available. What remains is a grudging acceptance, because the foresight a retreat would need is missing. The last word of the note is a warning: beware.

Wilhelm's line 2:

> Keeping his calves still.
> He cannot rescue him whom he follows.
> His heart is not glad.

His gloss treats the leg as unable to choose an independent path. It goes where the body goes. Stop the leg while the rest of the body is still driving, and the person falls. He maps that onto someone serving a stronger master. The servant may halt on a wrong road and still be unable to halt the other man. When that master pushes on, good intentions do not save him.

Three plots share the calf and nothing else. Hinton's threat is failed succession and failed contentment. Minford's other people cannot be harnessed, and the heart is heavy because retreat is closed. Wilhelm's calf belongs to a follower who cannot rescue the person ahead. Notes that say only "line 2: calves, unfavorable" have thrown the plots away.

Inside the "Bronze Age Oracle," the Zhou book with the later interpretations stripped off, Minford's name for Gen changes. It becomes "Tending." He thinks the hexagram points at older medicine, and at the need to look after the body. The Judgment for the whole figure:

> The back
> Is tended,
> The body
> Unprotected.
> He walks
> In an empty courtyard.
> No Harm.

He reads the empty courtyard as a figure for the body as a whole, left without care. Line 2 in that layer:

> The calves
> Are tended.
> There is
> No strength
> In the flesh.
> The heart
> Is sad.

The gloss stays physical. He hears calves carrying too little meat. Dropping weight is the worry, and it reaches the feelings directly.

Minford's two layers do not even share a name. Wisdom-layer Gen is "Mountain," a scene of stillness. Oracle-layer Gen is "Tending," a body looked after too little. Same six lines, retyped rather than merely annotated. An entry that prints only one of his names has dropped half the finding.

Arthur Waley and Richard J. Smith, whose short monograph on the *I Ching* sits in Princeton's Lives of Great Religious Books series, both carry Hexagram 52 back to sacrifice of humans and of animals, a widespread Shang custom. Smith's verb for Gen is "cleave." On a wholly separate reading, he says the word might instead mean to glare at something. His Judgment is puzzling, and the puzzlement is the point:

> If one cleaves the back he will not get hold of the body; if one goes into the courtyard he will not see the person. There will be no misfortune.

On line 2 he is blunt:

> Cleave the lower legs, but don't remove the bone marrow. His heart is not pleased.

Waley hears "gnawing," and he takes the hexagram to be about reading an omen from what small scavengers do. Rats, mice, and creatures of that sort meet the body of a sacrificial victim that has been put out so an ancestral spirit will come to it. His Judgment:

> If they have gnawed its back, but not possessed themselves of the body,
> It means that you will go to a man's house, but not find him at home.

Line 2:

> If they gnaw the calf of the leg, but don't pull out the bone marrow, their (i.e. the ancestors') hearts do not rejoice.

Put the verbs in one column, next to the people who chose them.

<table border="1" cellpadding="6" cellspacing="0" style="border-collapse:collapse;width:100%">
<thead><tr>
<th style="text-align:left;padding:6px;border:1px solid #ccc">Who prints it</th>
<th style="text-align:left;padding:6px;border:1px solid #ccc">Name or verb</th>
<th style="text-align:left;padding:6px;border:1px solid #ccc">Which layer they claim</th>
</tr></thead><tbody>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Minford, wisdom</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Mountain</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">later, still</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Hinton</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Stillness</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Taoist or Ch'an surface</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Wilhelm</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Keeping Still</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">his amalgam, mountain retained in the title</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Lynn, via Wang Bi</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Restraint</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Han comment</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Minford, oracle</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Tending</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Zhou reconstruction, medicine</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Smith</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">cleave, or else glare</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">sacrificial</td>
</tr>
<tr>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">Waley</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">gnawing</td>
<td style="padding:6px;border:1px solid #ccc;vertical-align:top">omen taken off a carcass</td>
</tr>
</tbody></table>

The figure did not change while the verb did. Mountain over mountain stayed mountain over mountain. The back stayed in the sentence. The courtyard stayed in the sentence, empty or merely unseeing. The calves stayed on line 2. What a subject does to the back, and what an empty court means, is the unstable field.

Some of the split is lexical: one Chinese word wide enough that "cleave" and "still" can both be proposed. Some is commentary, Wang Bi's picture of two people back to back. Some is the translator's own frame: nerves, a lesson about expectation, Latin, the yoga guess. "Translations differ" is true and useless. Naming which kind of difference you are looking at can be checked.

## What you can still assert once the face is gone

![A circular Shang bronze mirror whose reflecting face has vanished under heavy blue and green corrosion.](https://i.imgur.com/J9BMUU2.jpeg)

State the certain part narrowly. The narrowness is the result, not a failure of nerve. Hexagram 52 is two Mountain trigrams. It has to do with the back. It has to do with a courtyard, either vacant or occupied by people who are not seen. Outside that core, these few lines may concern keeping quiet, holding no expectations, restraining oneself, a settled mind, knowing when not to trail after a leader, tending aches and thin flesh, glaring, or the prep and the outcome of sacrifice, human or animal.

None of those readings has to be written off as a mistake or a bad translation. You could say the *I Ching* mirrors whatever the reader arrived hoping to see. Weinberger's correction of that saying is the better image, and it survives contact with the versions above. Think of a Shang bronze mirror. Manufacture and name still say mirror. A dark crust, blue going into green, has killed the reflection. People keep looking, and they keep reporting a face. The face is hard to separate from the hope they brought to the metal.

Minford remembers a last talk with David Hawkes, who was dying. Hawkes told him to pass on the secret: nearly any sentence can be taken in endlessly many ways, and nobody is ever going to know the sense it really has. That is the same property as the token that covers both a war captive and sincerity. No book has been altered as often as the Book of Change. Change is not only the subject. It is what happened each time a school, a dynasty, a missionary, or a poet bound a new specification to the same six marks.

When a new edition shows up, ask five questions of Hexagram 52 before trusting the rest. Whose name is printed: Mountain, Stillness, Keeping Still, Restraint, Tending, cleave, or gnawing? Is the Judgment kept apart from the Image? Are six line texts present, and can you find line 2 without reading the essay first? Is a commentator named when a commentator speaks? Does the page show the six lines, or only the English? A page that fails may still be a powerful essay. It is a poor specification of Gen.

## Run the cast, then attribute the string

Hexagram 52 teaches the larger point in miniature: this classic never froze into one meaning. Whoever consults it brings a question the pages do not contain. The coin procedure makes that input concrete, and it makes the encoding visible in a way commentary does not.

Ask in an open form. A demand for yes or no asks the lines to spell a Boolean they do not spell. Throw three coins, six times in all. Number the faces, add each throw, and let the total choose a broken mark or a solid one. Place the first result in the lowest slot and the sixth at the top. Flag the lines that change. Those flags carry the sharpest counsel. When the flagged lines invert, they address a second figure, the relating hexagram, read in the tradition as the direction things are moving.

![Three aged coins with square holes, caught mid-toss, beside a bundle of yarrow and a sheet marked in broken and solid strokes.](https://i.imgur.com/VZFkTgm.jpeg)

Coins plus a notebook are enough. Write the question before anything is thrown, then the six results, then which lines move, then the two names. Fifty yarrow stalks are the slower classical route to the same class of record: a count, not a toss, and more labor, not a different doctrine. Picking the quicker route does not pick a verb for Gen.

A page can own the arithmetic. [Free online three-coin casting](https://ichingonline.pro/) returns the figure you reached, each line that moves, and the relating figure, requires no account, and keeps the reading in the browser. Procedure can be faithful and the English can still be somebody's choice. Mountain stacked on mountain may arrive titled Keeping Still, Mountain, Stillness, Restraint, or Tending. Hawkes is not answered by a correct stack. Counting was never the step that produced "cleave" or "gnawing."

Store a line as a solid or broken mark, plus a flag if it moves. Store a hexagram as six ordered lines, line 1 lowest. Do not use the English name as a key. Keep Judgment, Image, line texts, the Ten Wings, and any translator's note in separate columns. Keep the rival verbs attributed, not averaged: Mountain, Stillness, Keeping Still, Restraint, Tending, cleave, gnawing. "The book says keep still" has dropped Wilhelm's name and kept his verb. Show the six lines with the result. Paz, Ginsberg, Borges, and Olson found beauty in his language, and Cage found a procedure. Those are real uses. They are not a substitute for the stack.

Leave the question in your own notes. The claim worth storing is not that a hexagram already contains a life. A figure reached under a question can be read against that question, and a different question puts the same Gen to different work. The uses never meet: a philosophical sorting of the universe, ethics, a ruler's handbook, an oracle for personal and political futures, then Christian proof, binary arithmetic, a peasant omen, a pseudo-science, Jungian chance, a Beat poem, a generative procedure. Weinberger's question, what the book is, stays open because those uses record reception, not a definition of the six lines.

Hold the split if you hold nothing else. Eight trigrams. Six lines. Sixty-four hexagrams. A cast with three coins or with fifty yarrow stalks, lowest slot first, moving lines opening onto a relating figure. On that small structure, an argument about what Gen does with a back, a courtyard, and the calves at line 2 has never closed. Mountain. Stillness. Keeping Still. Restraint. Tending. Cleave. Gnawing. Under the corrosion the Shang mirror is still a mirror by make and by name. It will not hand back one face.

A Developer's Field Notes on 10 Suno Alternatives (Tested 2026-10-18)

baiwei — Mon, 19 Oct 2026 04:10:22 +0000

Author: Baiwei.

I keep audio jobs in the same shape as any other integration: a fixture, an input contract, and a pass/fail note I can hand to someone else. Last Tuesday, around 11 p.m., that habit collided with a real deadline. A friend, Chen, sent a 40-second picture lock for a local coffee brand. The cut was due Friday. Music was not a line on the budget. The brief was three fields and nothing else: a female vocal, 92 BPM, and one Chinese line sitting in the chorus.

Suno was the first process I ran. Six generations. Take three had a melody I could cut against picture. The Chinese syllable in the chorus still sat wrong, and the UI would not let me patch that line. It only offered another full roll. Take four emptied the remaining credits. Friday did not move, and I still had no written answer on whether a generated master could ship inside a paid ad.

I closed the session and wrote down ten other stacks. The next morning I opened accounts and pointed the same fixture at each of them. Two weeks. Four tasks. On 2026-10-18 I read the price pages, quotas, and model names again, because those surfaces move faster than blog posts. These notes are the PR description I would paste if a teammate asked which generator to wire into a pipeline, and which license sentence they are actually buying.

I scored whether a file comes back, whether one bad section can be repaired, whether a DAW can open the pieces, and whether the plan you pay for is the plan that lets you publish. Vendor adjectives stay labeled as vendor adjectives. A commercial grant stays a license. It is not a copyright registration.

A 30-Second Spec for People Who Skip the Diff

If the standup is already over, these are the five calls I would defend from the fixture. The other five tools win on a narrower interface, and they get their own sections below.

Closest full-song replacement for a Suno-shaped pipeline: Mureka. Full songs, vocals, 12-track stems, a path into a DAW, and a real free entry.
Strongest edit surface on the model itself: ElevenMusic by ElevenLabs. Generations from 3 seconds to 5 minutes, inpainting, section composition, an audio reference with a copyright screen, and a finetune path.
Widest single workspace: Soundverse. More than one model, including Google's Lyria, plus stems, video, voice, and Agent One planning the calls.
Most surface area per budget dollar: AI Song Maker on aisongmaker.io, operated by CNINGOT LLC. The annual Basic seat is the cheap end of a very wide dashboard. Read the license before you treat the free column as a grant.
Fastest path from a finished song to a picture: Solmi. Upload an MP3 or WAV and the published render window is about 3–10 minutes.

Tunee, Loudly, ACE Studio, Musicfy, and One More Shot AI are not consolation prizes. They are different interfaces: a chat session that remembers you, a license paragraph you can paste into a statement of work, a MIDI vocal synth, a voice model, and a shot-level video desk. Pick from the ticket, not from the hero ranking.

Everything under this heading is the acceptance data behind those lines. Prices are the figures visible on 2026-10-18. They will change. Where a page renders a dynamic dollar amount, I do not freeze a number the checkout screen can contradict tomorrow.

Test Spec: Four Tasks, Ten Implementations

I treated the bake-off as a contract test. Same fixture. Same four assertions. For each tool I logged whether the step existed, how long the published or observed path took, and where it broke. A missing control is a fail even if the first listen was fine. A license page that disagrees with the pricing card is also a fail until you know which sentence you are shipping under.

Task	Input	Acceptance
Full song	A 30-second English pop chorus: female vocal, 92 BPM, one Chinese line in the chorus, plus one bridge so a single section can be targeted	Take one is usable against picture; generation time is recorded; one section can be repaired without a full reroll
Chinese delivery	Four Chinese lines plus two English lines	Tones land, breath is audible, phrasing does not collapse, and stray artifacts stay out of the vocal
Stems and post	Export vocal, drums, bass, and other; open the files in a DAW	The split is real, the session can import the format, and phase is either checked or explicitly left unchecked
Commercial path	The plan, credit rules, and DSP limits that apply on export day	The free tier is either usable for the ship or clearly not; the upgrade that unlocks publishing is named

Four filters sat in front of the heading list. A tool had to finish a song with vocals, not only a bed. After the song existed there had to be a next step: stems, a section edit, mastering, or video. Official pages had to be reconcilable, and a disagreement between two official pages counted as a finding rather than a reason to average them. Something real had to have shipped in the model stack during 2026.

One scope cut, so nobody files a bug against the list. SOUNDRAW is a solid fit when the ticket is royalty-free background music. It does not generate full vocal songs, so it is not one of these ten. If the only deliverable is an instrumental bed for YouTube, evaluate it on its own and do not force it through this vocal fixture.

No sample rate, unless the product page actually printed one. ElevenMusic is the stack that prints MP3 at 44.1 kHz, 128–192 kbps, or WAV. Everyone else gets an explicit "not published on the pages checked on 2026-10-18."
No phase-correlation score. The stem task asked about phase. Where a vendor only advertises a stem count, the count is the finding and phase stays unchecked.
No dollar amount for Loudly or One More Shot AI. Both hide a stable public rate card. Confirm inside the product.
No claim that a commercial license is a copyright you can register. Permission to use is not ownership, and it is not a promise that a registry or a third party will agree with you.

The friend's failed Suno pass is the reason the repair clause exists. A melody you cannot patch is an API that only supports POST /songs and not PATCH /sections. That is the bug this fixture is built to catch.

Mureka: Closest Drop-In for a Full Song Pipeline

Mureka started as a lyrics-to-song box. In 2026 the surface behaves more like a small production platform than a single prompt. The build under test is Mureka.

Model

The current model is V9.5, released on August 28, 2026. The release notes claim better song structure, more room in the arrangement, more natural vocals, and tighter prompt control. Treat those four adjectives as release-note claims. The numbers that can sit next to them are below. They come from a public run and from the vendor's own listener pass.

You can start from a prompt, from lyrics, or from a reference melody, then set genre, mood, tempo, key, and vocal style. That is the control surface I would map onto a form before I mapped it onto an API. After a song exists, the platform keeps going: Studio, Remix, 12-track stem separation, region editing, AI music videos, voice cloning, an API, and Mureka Co. Mureka Co is the piece that matters if you already live in a DAW. You generate parts and arrange the song with natural language instead of bouncing back to a browser tab for every change.

Duration

In a public six-brief, same-prompt test, V9.5 averaged 86 seconds per song and landed tracks around 3:17. In that same run it was about 1.9 times faster than MiniMax Music 3. That 86 seconds is the published same-prompt comparison, and the 3:17 figure is the length of the resulting tracks in that run, not a cap. If your pipeline assumes a 30-second chorus fixture, budget the public average as a queue time, then expect a full song rather than a loop the length of the prompt.

Stems

Twelve tracks is the published separation depth. The four-bus acceptance in this fixture was vocal, drums, bass, and other. Twelve is a superset of that bus if the export actually splits, which is why Mureka stays on the short list for anyone who has to reopen a session on Monday. Pro is the tier that adds stems, along with WAV and instrumentals. I do not have a published phase measurement for those twelve files. If phase cancels a vocal against the mix, you will find out in the DAW, not in the marketing page. Record that check yourself before you automate a bounce.

Sample rate and files

Base output is MP3. Pro adds WAV, instrumentals, and stems. A sample rate and a bit depth were not printed on the pages I re-read on 2026-10-18, so I am not going to stamp 44.1 kHz on the integration contract. If your downstream loudness stage assumes a specific rate, convert explicitly after export instead of guessing that the MP3 decoder and the WAV writer agree.

License boundary

There is a free entry. That is enough to run the 30-second chorus and see whether region editing can repair the Chinese line without burning a full reroll. It is not, by itself, a statement that the free file is cleared for a paid coffee ad. Plan names also disagree between official pages. The name you script against has to be the name on the checkout screen the day you subscribe, which in this pass was 2026-10-18. Whatever grant you do get is a license under their terms. It is not a registrable copyright, and it is not a warranty that a sync client will accept the file without questions.

Price on 2026-10-18

Basic is about $8 per month on annual billing. Pro is about $24 per month on annual billing. I would not hard-code either figure into a billing test that runs unattended next quarter. I would hard-code the date of the snapshot and fail the job if the checkout labels drift. Independent long-term user reviews are thinner than the feature pages. The roadmap moves fast, and the plan vocabulary moves with it. That is the operational risk, not the model number.

Fixture log

The public six-brief run is the timing evidence. Separately, Mureka's own evaluation put 20 listeners on 100 prompts. They called 61% of the generations a convincing lead vocal, with 97% prompt adherence. That is a vendor listener study with a small panel and a large prompt count. Useful as a prior. Not a substitute for your reference track.

Several Suno users report cleaner pronunciation and fewer stray humming artifacts. I am leaving that as a report from people who switched, not as a percentage I measured on the four-Chinese-line fixture. The fixture still wants tones, breath, phrasing, and a quiet vocal. Region editing is the control that maps to "fix the one bad chorus line." If that control is what your ticket needs, this is the shortest path from "I have an idea" to "I have a finished song I can still edit."

Open the same entry I tested, Mureka, and read the checkout names before you script a plan enum.

ElevenMusic: The Model With an Edit API in the UI

ElevenLabs launched Eleven Music on August 5, 2025. On September 11, 2026, Music v2.5 became the default model for prompt and reference generation on ElevenMusic. The product I linked for this pass is the ElevenLabs entry at ElevenMusic. There is a second consumer front door, ElevenMusic.io, with its own plan and its own license sentences. Do not merge them in a config file because the logos look related.

Model

Music v2.5 is the default you get for prompt and reference work. The changelog talks about layered arrangement, more natural instruments, and more depth. Those are vendor claims. The study they published next to the model is a blind comparison of 47,885 same-prompt pairs, and listeners preferred v2.5 in most of those pairs. It is a large paired preference from their evaluation, and it is the strongest published listener number in this entire set of notes.

The control list is the reason a developer stays. Natural-language generation. Vocals or instrumentals. Multilingual delivery. Composition plans. Section-by-section long-form writing. Inpainting. Genre changes in the middle of a track. Dense rap. Embedded sound effects. An API. Three of those matter more than the rest when the bug is "one line is wrong."

Audio reference takes about 30 seconds of audio. The system runs a copyright screen first and treats the clip as style guidance, not as a remix license. If the reference points at someone else's record, download is blocked. That is a guardrail, not a transfer of legal risk onto ElevenLabs. You still need the right to upload what you upload.

Inpainting is the PATCH. Highlight one bad chorus line, or one flat section, regenerate that span, and keep the rest of the timeline. This is the control Suno did not give me on Chen's spot, and it is the acceptance that separates a toy from a pipeline.

Music Finetunes trains a private style model on music you own. The published ready time is usually 5–10 minutes. That is a vendor timing for a finetune job, not a promise about your dataset. If you do not own the training audio, do not start the job.

Duration

A single generation runs from 3 seconds to 5 minutes. That range covers a sting, the 30-second chorus fixture, and a full short song without a second product. It does not cover an eight-minute marketing claim, and I am not going to stretch it. If the brief is longer than five minutes, you are in section composition, not in one shot.

Stems

The notes do not give ElevenMusic a stem count comparable to Mureka's 12 tracks or Loudly's nine. The post path that is actually specified is inpainting and section composition. If your ticket says "deliver vocal, drums, bass, and other as separate files," do not assume this model emits that zip because the editor is good. Confirm the export. I would rather fail the stem assertion in staging than discover it after a client asks for the drums alone.

Sample rate and files

This is the one page in the set that prints the audio contract. MP3 at 44.1 kHz, 128–192 kbps, or WAV. I would put those three tokens in the client docstring and reject anything else at the boundary. 44.1 kHz is a consumer rate, not a hidden high-resolution mode. If a broadcast spec asks for 48 kHz, resample on your side and document the conversion. Do not tell the client the model delivered 48 kHz.

License boundary

There are two entry points and two license systems. Free new songs on ElevenMusic can be used commercially if a visible "Made with ElevenMusic" credit ships with them. Only Pro new songs can drop that credit. Only Pro new songs can go to streaming platforms. Get the second sentence wrong and you will be reading terms after the release, which is the expensive direction. Attribution is a product flag, not a courtesy. DSP distribution is a different flag. A file can be "commercially usable" in a client video and still be the wrong plan for Spotify.

None of that language is a copyright registration. You are receiving permission under a plan. The screen that blocks references to other artists' songs reduces one failure mode. It does not assign you authorship, and it does not promise a collecting society will see the track the way the pricing page does.

Price on 2026-10-18

ElevenLabs plans start at $6 per month on the Starter tier. The separate consumer product, ElevenMusic.io, lists Pro at $9.99 per month, or about $7.99 per month if you pay annually, with up to about 1,000 songs per month. If you are embedding generation in your own app, you are almost certainly on the ElevenLabs side of that split, and the Starter price is the floor, not the invoice. If you are a listener-facing account that wants streaming rights without a credit line, you are on the ElevenMusic.io Pro sentence. Write both SKUs down. They are not aliases.

Fixture log

The time you save is in the section tools, not in raw generation speed. A bad chorus line gets a local regenerate. Reference audio goes through the copyright screen before it can teach the model bad habits or blocked material. Multilingual delivery is a stated capability. Run that lyric on the free tier before you bet a campaign on it.

Who it fits: producers and brands who will pay more than the cheapest annual seat on this list because they care about editability, consistency, and an API. The link I actually used is ElevenMusic. Read both license doors before you map a single plan field.

Soundverse: A Multi-Model Workstation With a Token Meter

Soundverse describes itself as "Music. Video. Voice. Directed by Agent One." For once the tagline matches the architecture. This is not a single generator with a skin. It is a workstation that can switch models while you work, and the comparison page names Google's Lyria among the models you can call. The account I opened is Soundverse.

Model

Agent One is the orchestrator. It decides which tools to call, then it asks you to confirm before it runs. In practice it asks two or three clarifying questions, and the first pass usually needs another round. That is a human-in-the-loop agent, not a fire-and-forget endpoint. If you are wrapping it, budget a confirm step. Do not write a worker that treats the first plan as the render.

The feature list is the widest in these notes: song generation, text-to-music, text-to-song, AI singing, lyrics, voice cloning, Artist DNA, similar-music search, looping, extending, remixing, mastering, vocal isolation and removal, music inpainting, a multi-track editor, music videos, and an API. Inpainting is on that list, so the "fix one line" acceptance has a named control, separate from simply chatting at Agent One until the whole song changes.

The rights stack is the other thing they advertise: TRACE watermarking and attribution, consent-based Artist DNA, and a clearance framework. I am logging those as company claims. They are not an independent legal opinion, and they are not a shortcut around the rule that you must own what you upload.

Duration

Song length is metered more clearly than it is capped. Text-to-music costs 1 token per second of audio. One extend costs 45 tokens. A chat message costs 1 token. A stem export costs 1 token. The free grant is 1,000 tokens, described as roughly three full songs. If you do the arithmetic on the published text-to-music rate, 1,000 tokens is 1,000 seconds of that meter before you spend anything on extends, chats, or stems. The budget comment I would leave in code is the vendor's own translation: about three full songs, then the meter is the source of truth.

Stems

Stem export is a priced action at 1 token, which is cheap next to an extend at 45. WAV and stems show up from the Creator tier upward. That is enough to satisfy the shape of the vocal-drums-bass-other fixture if the splitter actually emits those buses. A phase figure was not part of the published notes. The multi-track editor is the place you would hear a cancellation. Vocal isolation and vocal removal are separate tools from a full stem print. Do not collapse them into one capability flag in your UI.

Sample rate and files

WAV is the format named from Creator up. A sample rate and a bitrate were not on the pages checked on 2026-10-18. Same rule as everywhere except ElevenMusic: do not invent 44.1 kHz in the adapter. Convert downstream if the delivery spec demands it, and store the original file next to the converted one. The API is priced separately from the seat, so a prototype on Creator does not tell you the production invoice.

License boundary

Free-trial output cannot be monetized. The license also wants human-added content in the finished track. Those two sentences disqualify the free grant for Chen's paid coffee spot even if the chorus sounds right. You can evaluate the interface. You cannot ship the trial file as the ad. Read that again before you demo a client's logo over a free render.

The pricing page and the comparison page describe commercial rights differently. I am not picking a winner between two official sentences. That disagreement is the bug. Confirm the exact plan in writing before you publish, and screenshot both pages on the day you subscribe. A watermarking feature does not close the gap. TRACE is a claim about attribution, not a ruling that your use is cleared.

Price on 2026-10-18

Creator is $12 per month, about $10 per month on annual billing. Pro is $36 per month, $29 on annual. Max starts at $72 per month, $58 on annual for the entry tier of that band. Free remains 1,000 tokens. Token costs above are the ones that change a weekend prototype into an empty meter: the extend, not the stem. If Agent One likes to chat, remember that a message is 1 token. It is small until a loop of clarifications runs unattended.

Fixture log

The first session feels like a review, not a render. Two or three questions, a plan, a confirm, then a pass that usually wants a second round. That is acceptable if you wanted one workspace for the song, the stems, the vocal, the video, and the collaboration. It is the wrong tool if you wanted a single prompt box and a WAV three minutes later. Complexity is the cost of the width.

Soundverse does not have a separate published pronunciation score. Run the four Chinese lines and two English lines on the free tokens, and spend the rest of the grant on one inpainting pass so you know whether a bad tone can be patched.

The door I used is Soundverse. Confirm the commercial sentence on the plan you will actually be billed for.

Tunee: A Chat Agent That Remembers the Session

Tunee is not another form with a generate button. It is a chat-based music agent. You upload a melody, describe a mood, or hand it a video, then you iterate in the same conversation. The session keeps context and preferences, which is a real difference from a stateless prompt that forgets what you rejected three clicks ago. The entry I tested is Tunee.

Model

The product has grown from a chat box into a generative studio: a music agent, a song generator, AI characters, music videos, lip sync, AI dancing, and motion control. The mental model that fit the fixture is a producer who still remembers the note you sent three messages back. That memory is the feature. It is also the reason a failed instruction is frustrating. When the agent ignores a prompt, you cannot fall back on "the form fields were empty." The instruction was in the thread.

Tunee does not publish a model codename comparable to V9.5, Music v2.5, or MANTA 1. The interface under test is the agent plus the video tools, and the published listening pass is the four-week run below.

Duration

A public four-week test covered 37 songs across 12 genres and scored the tool 4.5/5. Chinese pronunciation in that run was clearly better than Suno's. That is the only head-to-head Chinese result in this whole document that I am willing to repeat as a comparison, and it belongs to that public test, not to a private panel I ran in a weekend. It is also the result that maps directly onto Chen's brief. A Chinese line in the chorus was the syllable Suno smeared. A tool with a published edge on Chinese delivery earns a longer look, with the caveats in the next paragraphs still attached.

Song-length caps were not printed as a hard ceiling in the notes I kept. Video length and video cost were the constraints that showed up in practice. Video burns credits quickly. If your worker renders a lip-sync preview on every chat turn, you will empty the day.

Stems

A stem count is not part of the Tunee notes I re-checked. If the ticket is vocal-drums-bass-other as files, this is not the tool I would pick first. The post path it does advertise is video: characters, lip sync, dancing, motion control. That is a different acceptance. Judge it against Solmi and One More Shot AI when the song already exists, and judge it against Mureka or ElevenMusic when the song does not.

Sample rate and files

No sample rate, bit depth, or container contract was stable enough in the checked pages for me to write one down. Treat the audio download as an opaque file until you inspect the header yourself. Do not document 44.1 kHz because a different product in this article prints that number. The free tier is about three songs a day, with attribution. That quota is large enough to A/B a chorus and a repair. It is not a batch API.

License boundary

Basic, Plus, and Pro grant commercial rights for qualifying content. Free output sits under a stricter set of rights, and the free songs want attribution. "Qualifying content" is doing real work in that sentence. It is not "everything the model emits." Read the qualifier on the plan you buy. A commercial grant on a paid plan is still a license, not ownership, and not a registrable copyright. If the qualifying rule excludes unmodified model dumps, a raw download may not be the thing you are allowed to sell.

Pricing transparency is the weakest of the ten. The full price table is not public. That means any integration doc you write has to say "confirm at checkout" instead of asserting a SKU table. Budget for a surprise when the card form loads. I would not let a procurement spreadsheet invent a middle tier to fill the hole.

Price on 2026-10-18

The official music-video page has referenced paid plans from about $18 per month. That is a reference point, not a complete rate card. I am not going to extrapolate Basic, Plus, and Pro into three neat prices the site refused to print. Free remains about three songs a day. If $18 is the floor you heard about, treat it as a rumor with a URL until the checkout shows a number on the day you pay.

Fixture log

The 4.5/5 and the Chinese-pronunciation note are the reasons to try it. The reasons to keep the credit card in your pocket during the trial are the Trustpilot pattern: failed lip sync, ignored prompts, and slow support. A chat agent that forgets an instruction is a flaky remote. A lip-sync fail after the credits are gone is a failed job you cannot cheaply retry. Support latency matters more here than on a tool whose only output is a WAV you can re-render yourself.

Who it fits: people who would rather talk than fill a form, and who want a short-form video from the same conversation without learning a DAW. It is a poor fit for a stem delivery or for a finance team that needs a public price URL in the ticket.

Try the same door I used, Tunee, on the free daily songs before you assume the paid video tools inherit the 4.5 score.

Loudly: Rights Language You Can Paste Into a SOW

Loudly used to be filed under background beds. That filing is stale. The stack now includes full songs with vocals, not only instrumental cues, and the production tools sit in the same dashboard as the generator. The entry point for this pass is Loudly.

Model

Instrumentals come from VEGA 1, VEGA 2, and VEGA 3. Full songs with lyrics and vocals come from MANTA 1. Sample generation can run through an integrated Stability 2.5 model. I am listing those as separate backends on purpose. The commercial sentence for Loudly's own output does not automatically wrap a partner model. If a render was produced by someone else's model inside this UI, the partner's terms can apply instead. Your adapter should store which model ID came back. A single provider=loudly tag is not enough for a license review.

Around the models: an AI remixer, a stem splitter, mastering, effects, a distribution workflow, an API, and SongDNA. SongDNA is the piece I would actually store. It analyzes audio, lyrics, structure, mood, tempo, key, and hooks, and it turns that into a fingerprint you can feed back into later prompts, captions, and cover art. That is closer to a content-addressed brief than to a chat log. If you are generating a series for one client, the fingerprint is how you stop the fifth cue from wandering off the first.

Duration

VEGA instrumentals usually come back in seconds. That is a practical observation about the instrumental models, not a benchmark suite with percentiles. MANTA 1 is the new vocal path. Complex genres occasionally repeat a pattern. Full-song vocals are the least proven part of the product, with less time in market than Suno, Mureka, or ElevenMusic. I would not put MANTA 1 on a flagship vocal spot without a longer audition than I would give V9.5 or Music v2.5. I would put VEGA on a bed much faster, because that is the older job and the free export is long enough to judge style.

The free tier only exports 30 seconds. That matches the length of the chorus fixture and fails the length of a release. You can decide whether the style is right. You cannot ship the free file as the master. No credit card is required to reach that 30-second export, which makes it a clean CI-style smoke test: create the account, render, confirm the cap, stop.

Stems

The splitter goes up to nine tracks. Nine is more than the four-bus acceptance and less than Mureka's twelve. It is enough to reopen a mix if the nine files are the stems you think they are. Phase was not published. Mastering and distribution live on the same dashboard, so the post path does not require a second vendor for a simple client delivery. Distribution inside the product is still not a statement about every DSP's AI policy. Read the destination's rules separately.

Sample rate and files

A printed sample rate was not part of the Loudly pages I froze on 2026-10-18. I will not copy ElevenMusic's 44.1 kHz into this row. Inspect the export. The fact that matters for the free tier is duration, not rate: 30 seconds, then a paywall. If your test asserts "file duration greater than 30 seconds" on a free token, it should fail. That failure is the product working as documented.

License boundary

Paid plans grant a worldwide, non-exclusive, royalty-free commercial license for Loudly's own output. Read the nouns. Worldwide, non-exclusive, royalty-free: that is a use right. It is not an assignment of copyright to you. It does not mean you own the composition. It does not mean you can register the track as if a human-authored work had been signed over. It does not bind a partner-model output that ships under different terms. For a freelancer putting music under a client video, a podcast, a game, or a piece of software, this is still the most explicit paperwork of the ten, provided you generated on Loudly's own models and you paid.

The free 30-second file is a listening tool. Do not paste it into a client cut and assume the paid grant applies backwards.

Price on 2026-10-18

The pricing page renders dynamically, and a stable official USD figure is not published in a form I am willing to hard-code. Old blog posts that quote 2024 prices are out of date. I am intentionally leaving the dollar cell blank. Confirm at checkout on the day you pay. If a teammate pastes a dollar amount from a screenshot into this doc later, check the date on the screenshot before you trust it.

Fixture log

Use the free tier, with no card, to hear VEGA in seconds and to confirm the 30-second wall. Use a paid plan only after you know whether you need MANTA 1 vocals or only beds. If the genre starts looping a pattern, that matches the note on complex material and MANTA 1 being new. It is not a reason to invent a failure rate. It is a reason to keep a human on the audition.

Who it fits: freelancers and studios who need music they can place in client work and who need a license paragraph that says what it grants. It is a weaker fit if the only question is "which vocal model has the most editing depth." That question still points at ElevenMusic.

The URL, unchanged, is Loudly. Check the model name on the render before you rely on the Loudly grant.

AI Song Maker: Aggregator Surface on aisongmaker.io

The name is a trap. This section is only about aisongmaker.io, operated by CNINGOT LLC in Sheridan, Wyoming. A different site, aisongmaker.ai, belongs to a different company. Do not mix their models, their prices, their support contacts, or their people. If a dependency review names the wrong domain, the rest of the license analysis is about a product you did not buy. The door for this product is AI Song Maker.

Model

The .io product behaves more like a multi-model aggregator than like a single in-house checkpoint. Public entry points include DiffRhythm, Riffusion, MusicGen, and ACE-Step. The commercial-license page cites the underlying open-source model licenses as part of the basis for allowing commercial use. That citation is a clue, not a comfort. Open-source model licenses are not identical to each other, and an aggregator page that points at them still has to be read. Store the model name next to the file. "Generated on AI Song Maker" is not a license identifier if one route was MusicGen and another was ACE-Step.

What the dashboard actually exposes is wide: an AI singer, a custom voice, VibeSeed, AI covers, vocal removal, MP3-to-MIDI, a MIDI editor, section replacement, photo-to-singing, and music videos. Section replacement is the repair control for the chorus-line bug. Lyrics can run to roughly 5,000 characters, which is far past the four-line Chinese fixture and long enough to paste a full lyric sheet without truncating mid-sentence. The marketing claim on duration is songs up to 8 minutes. I am labeling that as marketing. It is not a stopwatch result from the public six-brief style of test that Mureka has.

Duration

In the hands-on pass, the same lyric came back as a song quickly, and the dashboard had more entry points than you can scan in one look. The tool will carry songs past four minutes. The mix does not have the separation of the top tier. A third-party comparison puts it in the quick-prototype and budget bracket: fine for drafts, not the last stop before a commercial release. I agree with that placement as a workflow label. Drafts are a legitimate job. Shipping a paid ad from a draft model without a second pass is a different job.

Twenty daily credits buy about four generations. That is two rounds of comparison if you render two takes per round. It is not a load test. Do not point a cron at the free quota and expect it to survive a demo day.

Stems

Vocal removal, MP3-to-MIDI, and a MIDI editor are the post tools that are actually listed. I do not have a "12-track" or "9-track" number to put beside them. Vocal removal is not the same assertion as a phase-checked drum stem. If the client needs buses, plan on another tool after the draft, or verify an export this page did not document. Section replacement still lets you patch a span of the song, which is the repair half of the first task, even when the stem half is weak.

Music videos are a separate duration table: 10 seconds on Free, 2 minutes on Basic, 5 minutes on Standard and Pro. A 10-second free video is a storyboard sample. It is not a YouTube master. Do not let a demo of the free video set the client's expectation of length.

Sample rate and files

Outputs named on the plan are MP3, WAV, and MIDI. Sample rate and bitrate were not printed in the snapshot I trust. MIDI is the interesting one for a developer. If you can get a MIDI file out, ACE Studio or a conventional DAW can become the next stage without another full generation. That is a cleaner pipeline than hoping the mix separation improves inside the aggregator. Inspect the WAV header yourself before you promise a rate in a delivery spec.

License boundary

This is the sharp edge. The pricing card has placed a "Commercial License" mark in the Free column. The same page's FAQ says free users may only use songs for personal, non-commercial purposes. Those sentences cannot both be the contract. Trust the FAQ and the license page over the checkbox graphic. If you plan to sell, pay, and pay on the generation you intend to sell. A file you rendered on the free tier does not become a commercial file because you subscribed the next morning, unless the terms explicitly say so. Generate under the paid plan.

Even a paid commercial grant is permission, not ownership. Citing MusicGen or ACE-Step or the other open models does not hand you a registration. It tells you the vendor is leaning on those upstream licenses. Your review still has to match the route you used.

Price on 2026-10-18

Basic is $14.99 per month, or $10.49 per month on annual billing. Standard is $29.99 per month, or $20.99 annual. Pro is $59.99 per month, or $41.99 annual. The annual Basic figure is why this tool shows up in the "most features per dollar" line. The dollar is real. The license contradiction is also real. Cheap and unclear is not cheaper than a slightly higher plan with one sentence you can show a lawyer.

Fixture log

Use the 20 credits to run the chorus twice and a repair twice. Listen for mix separation, not for a miracle. If the draft is good enough to arrange and bad enough to mix, move the MIDI or the stems-you-don't-have problem to a DAW instead of buying the top tier for a separation the product did not demonstrate. Keep aisongmaker.ai out of the bookmark folder so a later you does not compare the wrong rate card.

Who it fits: budget-conscious builders who want many tools in one dashboard and who will read the license before they publish. The second copy of the URL is AI Song Maker. The FAQ wins over the free-column badge.

ACE Studio: MIDI In, Vocal Out, DAW Bridge

ACE Studio is not a one-prompt song generator. That is the point, and it is why an empty project stalls. The workflow is MIDI plus lyrics. You decide pitch, phrasing, vibrato, and emotion note by note. The model renders the singing. Musical decisions stay with you. If your repo has no melody, this is the wrong first dependency. The site I used is ACE Studio.

Model

The catalog lists 140 or more voices on some pages and 160 or more on others, across 8 languages. I am keeping the range, because the pages do not agree on a single integer and I will not average them into a fake precise count. Language coverage matters for the bilingual fixture. A voice that claims 8 languages still has to be heard on the four Chinese lines. A catalog count is not a pronunciation score.

Around that vocal core: singing voice cloning, a voice changer, AI instruments, Generative Kits, a stem splitter, vocal-to-MIDI with lyrics, a music enhancer, sound-effects generation, a video composer, and music videos. The bridge back to a real session is ACE Bridge. It connects through VST3, AU, and AAX, with ARA and tempo sync. For a developer who already lives in a DAW, that plugin list is the integration, not the website. Tempo sync is how the 92 BPM brief survives contact with picture. If the plugin drifts, the Chinese line will not land on the cut even when the phonemes are right.

Duration

Give it MIDI and lyrics and about ten minutes gets you a vocal with audible breath and vibrato detail. That ten minutes is hands-on time, not a queue time for a full arrangement. The six expression parameters can be automated one by one: Formant, Energy, Tension, Falsetto, Air, and Breath. That list is the actual control surface. If you only needed a prompt, you will feel the ten minutes as friction. If you needed the Chinese line to breathe on a specific syllable, you will feel them as the first tool in this article that lets you edit like a performance instead of like a reroll.

It does not write the melody for you. An empty project stalls. That is correct behavior for a synth, and it is a failed acceptance if your ticket said "invent the song." Put ACE Studio second in the pipeline when Mureka, ElevenMusic, Tunee, or even a human sketch has already produced a tune you can transcribe.

Stems

A stem splitter is part of the product, as is vocal-to-MIDI with lyrics. I do not have a published track count to set beside Mureka's twelve or Loudly's nine, so I will not invent "it exports four buses cleanly." The practical post path for this tool is the plugin: you are already inside the DAW, printing audio from a MIDI performance. Phase problems, if they exist, are the same problems you have with any stacked vocal print. An independent review of the product also says the Canvas still falls short of a full DAW, and complex arrangements go back to your main DAW. I would believe that before I would try to replace a session template with the Canvas. ACE Bridge is the grown-up route. The Canvas is a sketchpad.

Sample rate and files

The plugin formats are the published contract: VST3, AU, AAX, ARA. An audio sample rate for offline export was not the number I could pin on 2026-10-18. In a DAW this usually means "the session rate," which is your setting, not theirs. Do not write their name next to 96 kHz unless you set the session to 96 kHz. The vocal you print inherits the session. Document your session, not a guess about their renderer.

License boundary

You can start free. The free start is not a permanent full version. Do not architect a classroom image or a CI runner around an unmetered free seat that the vendor can close.

Prebuilt voices and instruments are described as royalty-free for streaming, monetized YouTube, ads, games, and client work. That is the vendor's description of a license to use those assets. It is not a statement that the resulting song is a copyright you can register in every country. Voice cloning is a separate gate: it depends on whether you hold the rights to that voice. A slider that can sound like a person is not permission to clone that person. Bring your own rights or use the prebuilt roster.

Price on 2026-10-18

Artist is about $16.58 per month on annual billing. Artist Pro is about $22 per month on annual billing. Rent-to-own means you keep access after roughly two years. I would model that as a lease that converts, and I would keep the "roughly" in the comment. Two years of Artist at the annual rate is the figure a producer can justify if the vocal prints are replacing session-singer revisions. It is a bad purchase if you wanted a prompt box for a single Friday delivery. Chen's coffee spot would not have started here. A series of films with a locked melody would.

Fixture log

The ten-minute vocal with breath and vibrato is the success case, and only after MIDI exists. The stall on an empty project is the failure case the sales page can hide. Expression automation is how you would chase the one Chinese syllable that Suno would only regenerate wholesale. That is a different kind of fix than ElevenMusic inpainting, and for a composer it is the better kind. For a marketer with no MIDI, it is a wall.

Who it fits: songwriters, producers, and game or film composers who already have musical intent and want the vocal performance under control. The link stays ACE Studio. Bring a melody or do not bring the project.

Musicfy: Voice Models, Not a Song Factory

Musicfy does a narrower job, and it does that job on purpose. The job is the voice. You pick a preset vocal that is described as royalty-free, you train a custom voice on your own recordings, or you transform a vocal you already have. Covers, vocal replacement, and other voice-led workflows sit on top. There is an API if you need to embed the voice step in another product. Songwriting is not the center of the product, which is why I do not rank it against Mureka as a Suno clone. The URL is Musicfy.

Model

A public 60-day test covered 200+ voice generations and 50+ songs, and it scored the product 4/5. Voice cloning was the strength. A few clean vocal takes were enough to train a model. That is the implementation note I would trust: the training set can be small, and it still has to be clean, and it still has to be yours. Text-to-song does work. Complex genres smear. Pages load slowly. None of those three sentences require me to pretend I heard a new genre matrix. They are the published 60-day results, and "smear" is a reason to keep complex harmonic material on a different generator and only borrow Musicfy for the voice.

The homepage, still in 2026, labels Stem Splitters as Coming Soon. I am taking the homepage literally. A roadmap badge is not a stem export. If your acceptance says vocal-drums-bass-other this week, Musicfy fails that assertion today, and a blog post that hopes otherwise is wrong.

Duration

The 60-day window and the counts (200-plus voices, 50-plus songs) are the scale of the public test, not a render-time SLA. What you should budget is ears-on time for the training takes, plus another pass when a complex genre smears and you have to decide whether to simplify the arrangement or move the song job elsewhere. Slow page loads are a product bug for anyone automating through the UI. If you have the API, prefer it over driving the slow pages. If you do not, do not put this UI on the critical path of a live demo.

Stems

Coming Soon means not in the build I would ship against. Do not write a client proposal that lists stem splitting as a Musicfy feature because a navigation item exists with a future label. Vocal replacement assumes you already have a song, or at least an instrumental, from somewhere else. That somewhere else can be Mureka, ElevenMusic, Loudly, or a human session. Musicfy is a stage, not the whole pipeline. Treat a missing stem splitter as a hard fail for the third task, not as a soft "maybe later."

Sample rate and files

No sample rate was published in the notes dated to this 2026-10-18 pass. Voice audio still has to meet whatever the trainer accepts. The practical spec I can write without inventing numbers: a few clean vocal takes, your own voice or a voice you have rights to, no background-music soup in the training file. If the takes are noisy, you are training the noise. The 4/5 score does not specify a sample rate either. Leave the header check to the file you actually download.

License boundary

The Professional page and the Studio page list a commercial license. The Starter page does not. That omission is the finding. Do not buy Starter, ship a client spot, and argue that the brand "supports commercial use" in the abstract. The plan page you paid for is the page that has to say it. A higher tier's sentence does not drip down unless the terms say it drips down.

The main terms page shows a last-updated date of September 2023. That is old for a product that handles voice cloning and commercial rights in 2026. Old terms are not automatically void. They are a review flag. Screenshot them. If the product added cloning features after the terms date, ask support which document governs, and keep the answer. I would not want to discover the governing clause during a takedown.

One more line, because the marketing invites a mistake. "Sing like your favorite artist" is a feature slogan. It is not permission to clone a real singer. The public 60-day notes do not overturn that. A custom model trained on your takes is a different fact from a model trained on a celebrity a cappella you do not own.

Commercial use, where the plan actually grants it, remains a license to use. It is not a copyright registration.

Price on 2026-10-18

Starter is about $8 per month on annual billing. Professional is $20 per month. Studio is $56 per month during the promotion that was current for this snapshot. The promotion price is not a permanent list price. Write "promo" next to $56 or you will under-forecast next quarter. Free entry exists, with limits. Use it to hear a preset voice and to see whether your training takes converge. Do not use it as the legal basis for a paid publish.

Fixture log

The voice stage passes when the takes are clean and the genre is not asking the song model to do the hard part. The song stage is optional and weaker, especially where complex material smears. The stem stage fails on current labeling. The commercial stage passes only on Professional and Studio, and only as a license, with a terms page that has not been refreshed since September 2023. That is a complete scorecard without any new listening claims.

Who it fits: creators making covers and vocal-led clips who already hold an instrumental or a finished song from another tool. It also fits a product team that wants a voice API and is willing to own the rights question in their own terms of service. Open Musicfy and read the Starter page and the Professional page side by side before you pick a tier. The absence of a sentence is data.

Solmi: MP3 to Lip-Synced Picture

Solmi answers a downstream ticket: the song exists, and the publisher asked for a video. It does not replace the generator that wrote the chorus. If you are still on the "fix the Chinese line" bug, finish that on a music tool, bounce a file, then come here. Upload MP3 or WAV. A Suno download is an acceptable input if you actually have the file. The account link is Solmi.

Model

You pick an aesthetic. Solmi detects lyrics, builds characters and scenes, and lip-syncs the vocal. The published language coverage is 14 languages, which is wide enough for the bilingual fixture if the lyric detector agrees with what you sang. Detection is not the same as pronunciation repair. If the Chinese line was wrong in the audio, the video will mouth the wrong line with confidence. Fix the audio first. A video model cannot inpaint a phoneme it only sees as a waveform.

Casting from your own photos is how you keep the same face across scenes. That feature has a rights twin: you need the right to use those photos and the right to depict those people. The tool will not invent that right for you.

Duration

The official estimate from upload to a finished video is roughly 3–10 minutes. In the pass that matches this write-up, the path was: upload a three-minute MP3, approve a storyboard, then render, inside that same 3–10 minute estimate. I am not tightening the estimate to a single number. Network, length, and aesthetic choice can move you inside the window. What I will not do is quote a 30-second render just because the music fixture was 30 seconds. The video job was described on a three-minute master.

Aspect ratios are part of the contract, and they are more specific than the audio tools managed to be. YouTube at 16:9. TikTok and Reels at 9:16. Spotify Canvas as an 8-second vertical. If your publisher is a Canvas loop, an 8-second spec is the acceptance, not a full-length MV. If your publisher is a vertical ad, 9:16 is the acceptance. Generate the one the placement needs. A 16:9 master cropped in an editor is how faces lose the lip sync you just paid for.

Stems

Not the job. Solmi consumes a stereo master. It does not replace a stem splitter, and I found no reason in the product notes to pretend it prints drums and bass as separate files. Keep the music tool in the pipeline for the stem task. Solmi starts when the master is done. Passing a project file in and hoping for buses is a category error.

Sample rate and files

Input is MP3 or WAV. Output specs that are actually printed are about the picture: free preview at 20 seconds, 480p, with a watermark; paid plans at full length, 720p or 1080p, without the watermark. No audio sample rate is published for a re-render because the audio is the file you uploaded. Preserve your master. Do not let a video export become the only copy of the song. A compressed video soundtrack is a terrible archive, and 480p preview audio is worse. Store the WAV next to the MP4 and treat the video as a picture deliverable.

Pasting a Suno link often fails. Solmi's own FAQ says so, and the reliable path is to download the file first and upload that. If you are automating this, do not build the integration on a share URL. Build it on a file upload. The link failure is a known edge, not an intermittent mystery you should retry fifty times.

License boundary

The site says paid plans include commercial licensing. That is a site statement about a video plan. It does not clear the song inside the video. If the master came from Suno, Mureka, ElevenMusic, or anywhere else, the music license still has to allow the video use. Solmi cannot grant rights in an input you did not have. A watermark-free 1080p export is a product entitlement. It is not a chain-of-title document.

The free tier is a 20-second, 480p preview with a watermark, no credit card, plus daily credits for limited songs or lyric videos. That preview is enough to judge whether the aesthetic and the mouth match. It is not a publishable ad. The watermark is the contract made visible.

The harder finding is corporate identity. The terms, the homepage, and the press kit do not disclose a legal entity name or a team. Treat Solmi as a product. The pages do not name a legal entity. Pricing also renders dynamically, so the numbers below are the ones visible on 2026-10-18 and still need a checkout check. If your compliance process requires a vendor legal name before a purchase order, this gap is a blocker, not a footnote.

Price on 2026-10-18

Plus is $35 per month. Creator is $85 per month. Pro is $135 per month. Paid output is the full-length 720p or 1080p file without a watermark. I would not annualize these into a discount the page did not offer in the snapshot. Confirm the dynamic price at checkout, then screenshot it. The gap between a 20-second watermarked preview and a $35 plan is the real purchase decision. Make it after you have seen the mouth on your own chorus, not after you have seen a demo reel.

Fixture log

Three minutes of audio, a storyboard approval, a render inside 3–10 minutes, lyric detection across the 14-language claim, and a stable face when the cast comes from your photos. The Suno-link failure mode is specific enough to write a test for: a pasted link should be treated as an expected fail, and the file upload is the happy path. Udio users are in the same bucket as Suno users here. Bring a download, not a player URL.

Who it fits: anyone who already has the song and needs a vertical or a widescreen today, and who can tolerate an unverified legal entity on the vendor side. Speed is the feature. Title is the risk. Use Solmi for the upload test, and keep the master audio in your own bucket.

One More Shot AI: Shot-Level Video After the Song Exists

One More Shot AI aims at independent artists and labels. It is a downstream tool, and that is a compliment, not a slight. Do not ask it to write the chorus Suno missed. Compare its pictures to Runway, Kaiber, and Solmi. The song is an input. The control over shots is the product. I opened One More Shot AI.

Model

Three modes cover three levels of control, and they should be three code paths, not one "make video" button.

AI Artist builds a persistent virtual artist from a photo or a description and reuses that identity across videos. Persistence is the point. A one-off face that changes every render will not survive a three-video release schedule.

Auto Videos takes a song and photos, then AI agents arrange the video the way a small crew would. You trade control for speed. The published limit inside this mode is real: no model picker, and no per-scene storyboard. If you need to approve every shot before tokens burn, Auto is the wrong mode. Switch to Project Mode.

Project Mode is where you edit individual shots. That is the precision path, and it is why a label would sit here instead of on a one-click MV tool. It is also the path that costs operator time. Do not quote Auto's render clock as the cost of Project Mode.

The system analyzes waveform, beats, tempo changes, vocal segments, and mood to build synchronized visuals. There are Business, Labels, and API entry points for anyone who needs this outside a single artist login. The existence of an API entry is the finding. The schema is something you read after you are in the account.

Official showcase pieces run 2:26 and 3:19. Those are full videos, not five-second demos. I am citing the showcase lengths as published durations, not as the length of Chen's 40-second picture lock.

Duration

A four-minute song in Auto mode took about 20–30 minutes and produced 38 shots, with a consistent AI artist and mostly accurate lip sync. The on-screen countdown said 60 minutes and finished early. I am keeping both numbers because they teach different things. The 20–30 minute range is the observed job length for a four-minute master. The 60-minute countdown is a UI estimate that was pessimistic. If you are writing a progress bar on top of their progress bar, do not tell the user "60 minutes" as a promise, and do not tell them "20 minutes" as a promise either. Show their countdown, and log the actual completion.

Before you spend render tokens, a moodboard previews palette, story, and character stills. That preview is the cheap assert. Use it. Skipping it because Auto feels automatic is how you pay for 38 shots of the wrong wardrobe.

A four-minute song is not the 30-second chorus fixture. Scale your expectation. I am not dividing 25 minutes by four and declaring a per-minute SLA. The data point is one published job shape: about four minutes of music, Auto, 38 shots, 20–30 minutes of wall clock, countdown reading 60.

Stems

Out of scope. The input is the song plus photos. Shot edits happen in Project Mode. Nothing in the notes says you get vocal, drums, bass, and other back out. Do the stem work before the video work. A 38-shot render of a mix you cannot reopen is a painful way to discover that the bass was wrong.

Sample rate and files

Audio format specs were not the published contract. Picture and time were. You upload a song you already rendered, so the sample rate is whatever you bounced. Lip-sync quality follows the vocal clarity of that bounce. A chorus with a smeared Chinese line will sync to the smear. The countdown behavior is the other interface contract worth testing: the number on screen can overestimate, and finishing early is not a crash. Assert on job state, not on the first countdown value.

License boundary

You have to hold the rights to every song, every photo, and every person you upload. That sentence is broader than "don't clone a celebrity voice." A band photo, a friend's face, a sample you do not own, a Suno master your plan does not allow you to visualize: all of those fail the input test even if the renderer is happy to try. The tool will produce the frames. The problem arrives later, from a person or a rightsholder. Build the check before the upload, not after the render.

There is a free start. Pricing itself is shown inside the app. I am not writing a dollar figure, because the public page did not give me a stable one and the fact pattern says not to invent one. Confirm at checkout. A free start is enough to see the moodboard and to learn which mode you actually needed.

The company behind the product is KRNL SRL in Milan. The CEO title belongs to KRNL, not to a separately verified company named One More Shot AI. If a vendor form asks for the legal entity, write KRNL SRL, not a guessed brand entity. I am not going to invent a registration number or a share structure that was not in the pages.

Commercial use of the output still cannot outrun the rights in the input. And a platform license, if your paid plan includes one, is permission. It is not a copyright registration in the video or in the song.

Price on 2026-10-18

App-side pricing only. No USD number in this note. If a reseller page quotes a monthly dollar amount, it is not part of this snapshot unless you saw it inside the app on 2026-10-18 and saved the screenshot. I did not freeze one, so the cell stays "confirm in the app."

Fixture log

The success case is a consistent identity across 38 shots, mostly accurate lip sync, a moodboard you can reject for free, and a job that beat a pessimistic countdown. The failure case is Auto when you needed a per-scene storyboard, or any upload you did not have rights to. Showcase lengths of 2:26 and 3:19 are evidence that the tool is trying to finish real videos. They are not evidence about musical quality, because musical quality was not the task.

Who it fits: artists and small labels who want a stable visual identity and shot-level control, and who already have a master they are allowed to picture. For a same-day MV with less control, Solmi is the faster comparison. For the song itself, neither one is the Suno replacement.

The second link, same string, is One More Shot AI. Check the mode before you spend the render.

Side-by-Side on 2026-10-18

All paid-from figures below are the official numbers that were visible on October 18, 2026. They change. Loudly and One More Shot AI stay unpriced on purpose. A blank cell is more honest than a stale blog number.

#	Tool	Best for	Free tier	Paid from
1	Mureka	Full song replacement	Yes	~$8/mo annual
2	ElevenMusic	Model quality and editing	Yes, limited	$6/mo (ElevenLabs); $9.99/mo (ElevenMusic.io)
3	Soundverse	All-in-one studio	~3 full songs	$12/mo ($10 annual)
4	Tunee	Conversational agent plus video	Yes, stricter rights	Full price list not published; ~$18/mo referenced
5	Loudly	Rights-cleared production	Yes, no card	Dynamic pricing; confirm at checkout
6	AI Song Maker	Features per dollar	20 credits/day	$14.99/mo ($10.49 annual)
7	ACE Studio	Pro vocal synthesis	Limited start	~$16.58/mo annual
8	Musicfy	AI voice and covers	Yes	~$8/mo annual
9	Solmi	Song to MV, fast	20s watermarked preview	$35/mo
10	One More Shot AI	Controlled MV production	Yes	App-side pricing

How I would read this table in a design review. The "best for" column is a job, not a quality crown. The free column tells you whether the fixture can run without a card. The paid column is a 2026-10-18 snapshot, and two of the rows refuse to be snapshotted. If a teammate sorts by price and drops Loudly for having no number, they sorted a missing value into "expensive" or into "free." Both sorts are wrong. Open the checkout.

A second reading, against the four assertions. Full-song repair wants Mureka's region edit or ElevenMusic's inpainting, with Tunee's chat and AI Song Maker's section replacement as alternate patches, and ACE Studio as the MIDI patch. Chinese delivery has one published head-to-head worth citing: Tunee's public test against Suno. Everyone else still owes you a run of the four Chinese lines. Stems are real, today, on Mureka (12, Pro), Loudly (up to 9), and Soundverse (from Creator). ACE Studio has a splitter inside a vocal synth workflow. Musicfy's splitter is still Coming Soon. Solmi and One More Shot AI do not pretend. Commercial clarity is best on Loudly's paid own-model grant, sharpest as a trap on AI Song Maker's free column, and split across two products on ElevenMusic. Soundverse's two official pages disagree. Solmi's paid plans say they include commercial licensing, and the vendor entity is undisclosed. That is the whole table in prose. The grid is only an index.

Picking a Tool From the Ticket, Not the Homepage

Match the ticket to one line, then run the free tier that line points at. Do not subscribe to a stack ranking.

You want the closest one-to-one replacement for a Suno prompt-to-song flow, and you still want stems and a DAW path afterward. Start with Mureka. The public timing prior is 86 seconds and tracks around 3:17, the stem depth is 12 on the paid stem tier, and region editing is the repair.

You care about editing one bad line more than you care about the cheapest seat. Go to ElevenMusic. Inpainting and section composition save more time than a faster first render. Remember the fork: ElevenLabs Starter at $6 per month is not the same license as ElevenMusic.io Pro at $9.99, or about $7.99 annual, and only Pro new songs drop the "Made with ElevenMusic" credit and go to streaming.

You want one place for the song, the stems, the voice, and the video, and you can tolerate an agent that asks questions. Soundverse is the widest workspace here. Budget tokens, not just the seat. Creator at $12, Pro at $36, Max from $72, free grant roughly three songs, extend at 45 tokens.

You would rather chat than fill out a form, and Chinese delivery is on the brief. Tunee is built around the conversation and a public 4.5/5 across 37 songs. The price table is the part you cannot review in public. Assume a checkout surprise, and assume video spends the day budget.

You need music in client work and you need the paperwork to say "worldwide, non-exclusive, royalty-free" for the vendor's own output. Loudly. Keep VEGA and MANTA 1 distinct. Keep partner models out from under that sentence. Do not type a dollar amount until the live pricing page shows you one.

You have about ten dollars a month on an annual Basic and a long list of buttons. AI Song Maker on aisongmaker.io gives you the most surface per dollar, from $10.49 annual Basic up to Pro at $41.99 annual. Read the FAQ before you believe the free-column commercial badge. Remember CNINGOT LLC, and remember that aisongmaker.ai is not this vendor.

You already have a melody or MIDI, and the performance has to be directed. ACE Studio will beat every prompt box on vocal control. Budget about ten minutes and the six expression curves. Artist is about $16.58 per month annual. Do not arrive with an empty arrangement.

You need a singer, not a song. Musicfy. Professional at $20 and Studio at $56 promo list a commercial license. Starter at about $8 annual does not list one on its page. Stems are still Coming Soon. The terms date is September 2023.

You already have the song and you need a video before the end of the day. Solmi for speed: 3–10 minutes, 14 languages, 16:9, 9:16, 8-second Canvas, Plus at $35. One More Shot AI for shot-level control: Auto in about 20–30 minutes for a four-minute song, 38 shots, Project Mode when Auto is too blunt, price inside the app, entity KRNL SRL.

If two lines match, open both free tiers and run the same 30-second chorus. The one that lets you fix a single line without a full reroll is the one worth paying for. That test is cheaper than a month on the wrong seat.

Three Commercial Rules Before You Merge

This is the section listicles skip. It is also the section that shows up as a refund, a takedown, or a client who will not pay.

Rule 1: "Commercial use" is permission, not ownership.

When a platform grants commercial rights, it is giving you a license under its terms. The license is usually non-exclusive. Loudly says this in the clearest words of the group: worldwide, non-exclusive, royalty-free, for its own output. That wording is a use right. It is not a guarantee that you can register a copyright in every country. It is not a promise that no third party will ever object. It is not an assignment of the composition to your company. Put the license in the deal folder. Do not put "we own the master" in the deal folder unless a lawyer has told you, in writing, that this specific output supports that sentence. No tool in this list gave me that sentence.

Rule 2: Free tiers are a different contract.

AI Song Maker's own pages disagree about whether free songs are commercial. The card has worn a Commercial License badge in the Free column. The FAQ says personal, non-commercial only. Believe the FAQ and the license page. ElevenMusic lets free new songs be used commercially and requires a visible "Made with ElevenMusic" credit, while only Pro new songs can go to streaming platforms without that credit. Soundverse's free-trial output cannot be monetized, and the finished track is supposed to contain human-added content. Tunee's free output is more restricted than Basic, Plus, and Pro. Musicfy's Starter page does not list the commercial sentence that Professional and Studio list. Solmi's free file is a 20-second watermarked preview. Read the plan you will actually publish under. A trial render is a listening file until the plan says otherwise.

Rule 3: Input rights are your problem.

Upload a reference track, a real singer's voice, a melody you do not own, or a person's photo, and you need the rights to that input. ElevenMusic's copyright screen reduces one risk and then blocks downloads that reference other artists' songs. It does not move the risk onto ElevenLabs. Musicfy's cloning tools do not become legal because the button exists. ACE Studio's cloned voices depend on your rights. One More Shot AI requires rights in every song, photo, and person. Solmi will lip-sync whatever master you had the right to upload, and it will also lip-sync a master you did not. The renderer is not the rights department.

One habit that costs nothing: screenshot the pricing page and the license page on the day you subscribe. Plan names and prices move faster than articles do. The date on this note is 2026-10-18. If you are reading it later, the screenshots you take yourself outrank the table above.

FAQ

Is Suno still worth paying for in 2026?

Yes, if what you need is speed and the prompt-to-song feel. I burned a session on it for Chen's spot and the wall I hit was repair plus a credit cap, not "the model cannot sing." The alternatives in this note win on editing depth, stem workflows, clearer licensing, or video. Switch when you hit a specific wall. Do not switch because a ranked list told you to. If Suno already fixes the section you care about on your current plan, stay. If it only offers a full reroll, the first two sections of this note are the migration.

Which alternative has the most usable free plan?

"Usable" means "enough to judge the interface," not "enough to score a commercial campaign." Soundverse gives you roughly three full songs on 1,000 tokens. AI Song Maker gives 20 credits a day, about four generations. ElevenMusic lets you create free songs if you can live with attribution. Solmi gives a 20-second watermarked preview at 480p and does not ask for a card. Loudly lets you hear models without a card and then stops the export at 30 seconds. Tunee is about three songs a day with stricter rights. All of those are enough to run the chorus fixture. None of them are a blanket commercial grant. Mureka, ACE Studio, Musicfy, and One More Shot AI also let you start free, with the limits already spelled out in their sections.

Can I release AI-generated songs on Spotify?

It depends on the plan, not on the mere fact that a website generated audio. ElevenMusic is the explicit example: DSP distribution is limited to Pro new songs. A free song that is commercially usable with a "Made with ElevenMusic" credit is still the wrong file for a distributor upload. Other tools vary, and several never mention streaming by name. Check the distribution clause before you upload. A Loudly use-license, a Musicfy Professional license, or a Solmi video license does not rewrite a distributor's content policy. Read both documents. The expensive failure is a release week takedown, not a rejected preview.

Which one sounds the most like Suno?

Mureka and ElevenMusic are the closest in workflow and in full-song output. Mureka, in the public six-brief same-prompt test, averaged tracks around 3:17 with generations around 86 seconds. ElevenMusic's v2.5 won most pairs in the vendor's 47,885-pair blind evaluation. Taste is still subjective. Run your own reference through the free tiers, including the Chinese chorus line, and keep the take you can repair. If the ticket is a voice rather than a song, Musicfy is the closer comparison and Suno is the wrong baseline. If the ticket is a video, you are comparing Solmi and One More Shot AI to each other, not to Suno.

Do I have to credit the tool?

Only when the plan says so. ElevenMusic's free tier requires the visible credit "Made with ElevenMusic." Pro new songs do not. Tunee's free tier expects attribution, and the paid Basic, Plus, and Pro grants are the ones that talk about commercial rights for qualifying content. Other products vary by plan, and a few hide the answer in a FAQ that contradicts a pricing badge. If the credit is required, it is part of the deliverable, like a font license. Ship it where the audience can see it. Do not bury it in a README and call the ad clean.

Bottom Line

The best Suno alternative in 2026 is whichever one does the job after the song exists. Pick Mureka when you want the closest replacement and a 12-track stem path. Pick ElevenMusic when you need to patch one line and you can keep the ElevenLabs plan and the ElevenMusic.io plan straight. Pick Soundverse when the workspace has to cover song, stems, voice, and video, and you will watch the token meter. Pick Solmi when the master is done and you need a picture in 3–10 minutes. Pick One More Shot AI when the picture needs shot control and you can accept app-side pricing from KRNL SRL.

Leave SOUNDRAW off this list unless the job is a royalty-free bed with no lead vocal. Leave aisongmaker.ai off the invoice unless you meant that other company. Do not write a dollar price for Loudly or for One More Shot AI until the live screen shows you one. Do not tell a client that a commercial toggle is a copyright registration.

Then do one thing today. Open two free tiers. Run the same 30-second chorus, female vocal, 92 BPM, one Chinese line. Try to fix that line without regenerating the whole file. The tool that lets you fix it is the one worth paying for on a snapshot dated 2026-10-18.

Some links in this article are affiliate links, and the author may earn a commission.

Detect, inpaint, match style, then render: a developer’s model of editing lettering in a flattened image

baiwei — Fri, 09 Oct 2026 14:00:04 +0000

A flattened image does not contain a text node. It contains pixels that happen to look like letters. The poster, the menu photo, the label, the screenshot, the ad export: the type layer is gone, the style run is gone, and the only API you have is the raster. If you are choosing or building a feature that lets someone type a replacement and expects those pixels to become different letters without looking pasted on, you are not calling one model. You are running a stack, and the stack fails at its weakest stage.

The marketing sentence is always the same. Upload a picture, type a change, and the software erases the old words and draws new ones that blend in. That sentence hides a contract between stages. Detection can box the wrong strokes. The hole can be filled with a texture that does not belong. The style read can guess a family the renderer does not have. The new glyphs can be blurry, misaligned, or misspelled. Any one of those is enough to fail a human eye, which is the only judge that matters for a finished design.

This is the developer cut of that problem. I walk the stages the way you would sketch them before writing an interface: what each stage consumes, what it must emit, and how a bad emit poisons the next call. Then I write down the comparison harness, including the metrics I planned and did not publish. Then a matrix of the products that claim one or more of the stages. Then who should call which camp. I am not pasting a scoreboard. Where a metric was designed and not computed, I say the plan and withhold the number.

The practical gap is the one you already suspect if you have shipped a demo. A recorded demo picks a clean word on a flat color and stops the video before the proofread. A real job is a price, a date, a translated dish name, or a slogan on a photograph. The demo and the job are not the same function. Treat them as different tests or you will approve a pipeline that only works on the demo distribution.

The stack, stage by stage

Editing lettering in a flattened image is not one task. Treat the work as four stacked jobs — detect and OCR, inpaint, style match, render new text — because the composite is only as good as the weakest job. The data flow I want on a whiteboard is short enough to implement against:

input image → text detection and OCR → font and style analysis plus an edit mask → AI content fill, which is the inpaint → new text render with the style preserved → composite → edited image.

Most mainstream products stitch recognition, a fill, and a text update into that shape. The maturity of each module, and whether they share a mask and a string instead of each inventing their own, decides whether the edit survives contact with a person who can read. A stage that cannot name its input is not a stage. It is a prompt with a hope attached.

Here is the contract I would actually code between those jobs. None of these fields is optional if you want the run to be debuggable.

Box:
  quad
  raw_string
  confidence
Style:
  family_guess
  size
  color
  stroke
  shadow
  warp
Mask:
  hole
Edit:
  target_string
  operation

operation is either a replacement or a deletion. Deletion stops after the fill. Replacement continues into the renderer. If your product collapses all of that into one prompt string, you have not removed the jobs. You have made them impossible to unit test.

A second contract sits beside the schema: the source bytes are immutable. Write the edited raster to a new object. Keep the boxes, the mask, and the target string in a sidecar so a later pass can re-render without guessing what the model thought it saw. When a result looks wrong, you want to know which field was wrong. You cannot get that from a single flattened PNG and a shrug.

Detect and OCR

The first job finds the regions and reads them. Modern recognition combines deep learning, typically a convolutional front end with an LSTM sequence model, or a successor that does the same split of vision features and character order. On printed type with a clean ground it is solid. It is weak on low resolution, tiny fonts, handwriting, and busy backgrounds. That is not a footnote. Accurate detection is the foundation. Get the box or the string wrong and every downstream job fails in a way that looks like a rendering bug.

Think about what you store. An axis-aligned rectangle is the wrong primitive for a rotated sign or a word on a curved bottle. The rectangle includes background and misses glyph corners, so the mask is already dirty before the fill model runs. A quad, or a tighter polygon around the strokes, is the honest geometry. Multi-line grouping matters too. A headline and a subline that share a color are not one string, and a single box around both will ask the renderer to paint a paragraph into a title’s footprint.

Confidence is a gate, not a decoration. If the reader is unsure, the interface should show the raw string and let a person correct it before anyone inpaints. Silent failure here is expensive: the user “fixes” a word the model misread, the fill erases the real word, and the render dutifully draws the wrong correction in a beautiful font. Dedicated editors start by auto-boxing regions so a person can select one and edit it directly. That is the right product shape. Detection proposes. A human disposes. Fully automatic boxing with no confirm step is how you ship a tool that is confident and wrong.

Low resolution is a different failure from a tiny font. A tiny font at a high raster resolution may still have enough pixels per stem. A normal font photographed from across a street, then compressed, does not. Busy backgrounds break both the detector and the reader because edges of the artwork compete with edges of the strokes. Handwriting breaks the language model inside the reader, which was trained on type. None of these are reasons to skip the confirm step. They are reasons the confirm step exists.

There is also a product decision hiding in the box. Do you let the user edit only what was detected, or do you let them paint an extra region the detector missed? The second path is mandatory. Detectors miss reversed type, type in a shadow, and type that is the same color as a gradient behind it. A pipeline with no manual box is a pipeline that cannot recover. A pipeline that only has a manual box, and never proposes one, pushes the segmentation problem onto every user, which is the other failure mode. You want both: a proposal, and a brush that can add or subtract.

The string you show the user should be the string you will erase. Do not run a second, secret recognition pass later and erase a different set of glyphs. I have seen that class of bug in other document tools: the preview says one thing, the mask was built from another hypothesis, and a leftover stroke remains because the two hypotheses disagreed by a character. Pin the hypothesis. Persist it. Erase that.

Inpaint the hole

Once the mask exists, something has to invent the pixels that were behind the glyphs. Older inpainting such as PatchMatch copies nearby patches by nearest-neighbor search. It is fast, and it looks fine when the surround is stationary: a flat wall, a simple gradient, a repeating weave. It fails when the hole is large, when the texture is not stationary, or when a lighting falloff crosses the word. The copied patches tile, the lighting breaks, and you can still read the shape of the old word as a scar.

Those exemplar methods have largely been replaced by generative models, diffusion-based or GAN-based, that condition on the unmasked context and try to keep lighting and texture continuous. That is a real upgrade for photographs. It is also a new way to be wrong. A generative fill can invent a plausible brick pattern that was never there, or continue a highlight through the hole in a way that bends the light. For lettering removal, “plausible” is not the spec. The spec is “the surface that was already there, with the ink gone.”

Mask thickness is an engineering parameter, not a default you forget. Too tight, and a halo or a partial stroke remains. The next stage will sample that halo and treat it as part of the style. Too loose, and the fill eats a logo edge, a face, a rule line, or the neighboring word you meant to keep. Dilate and feather on purpose, and log the kernel you used. A removal demo will often describe this step as reading the surround and rebuilding whatever sat behind the glyphs. That description is the right mental model even when the implementation is a latent fill rather than a patch copy. Judge the result by whether a person can find the old baseline, not by whether the hole is “creative.”

Lighting gradients and texture are where this stage earns its keep or gives up. A word on a wrinkled cloth, a word on foliage, a word on a chrome edge: the conditional model has to continue a non-repeating signal through a hole shaped like letters. Letter-shaped holes are adversarial. They are thin, they have high-frequency boundaries, and they sit on top of the very signal you wish had been observed. Expect scars there. A flat printed sheet does not ask for this heroics, which is why so many demos choose it.

If you only delete, you are done after a successful fill. If you replace, the fill is an intermediate, not the product. It has to leave a surface the renderer can sit on. A fill that still contains ghost strokes will show through antialiased glyphs. A fill that changed the local noise profile will make the new word look like a sticker even when the font is right. Look at the hole before you look at the new word. Developers who only screenshot the final composite debug the wrong stage.

One more implementation note. Run the fill on the original pixels, not on a recompressed preview. Compression blocks along the mask boundary become part of the context the model trusts. If your UI shows a small proxy and your server fills the full raster, log both sizes in the sidecar or you will “fix” a bug that was only a proxy artifact. The same rule applies when the client downscales for upload and the server never sees the stems of the letters. You cannot inpaint detail you already threw away. Prefer a crop of the region plus a ring of context over a blind full-frame downscale, and keep the crop coordinates so the composite lands back on the original grid.

Match style before you draw

New lettering should match the original’s typeface, color, shadow, stroke, and distortion. That is its own stage, not a side effect of the fill. You are trying to separate a word’s content from its style, then apply the style to a string the user just typed. If you skip the separation, the renderer either uses a default font or asks a generative model to remember the font, and generative models are bad at remembering fonts.

Meta Research's TextStyleBrush extracts a text style — font, stroke weight, warp — from a single example and applies it to new content. That is one-shot style transfer. It is a research prototype, not a product in this test, but it shows the split is real: content in one vector, appearance in another. Commercial tools approximate the split with a built-in font match for a similar family, size, and color. Clipfly, for example, advertises editing lettering in the same font format. Approximation is the right word. A match against a library is not the same as a shader that reproduces a custom logo outline, a warped baseline, or a hand-painted terminal.

What the style stage must emit, if you want the renderer to be dumb and predictable:

A family guess that maps to a font you can actually rasterize, plus a flag when the guess is weak.
Size in pixels, measured on the quad, not on an axis-aligned box that inflated the height.
Fill color and an optional stroke, sampled from the glyphs, not from the halo you forgot to dilate away.
A shadow or extrusion description, even if the description is “none.”
A warp or perspective mapped from the quad, so a word on a receding sign does not come back as a front-facing sticker.

If any of those fields is missing, do not pretend the render is style-preserving. Say the match failed and fall back to a visible default, or stop and ask. A silent fallback to a popular grotesque is how “AI edited my poster” becomes “AI replaced my brand.”

Length is part of style even though it feels like layout. The user’s new string is rarely the same width as the old one. The stage that owns style has to say whether the renderer may scale to fit, track tighter, wrap, or refuse. Scaling to fit changes the weight and breaks the match you just measured. Tracking tighter can collide with the decoration you did not mask. Refusal is a valid output. A pipeline that always squeezes is a pipeline that lies about fidelity.

Color sampling has its own trap. Glyphs are antialiased, so a naive average of the bounding box mixes ink with background and returns a muddy color. Sample the cores of the strokes, or the darkest or lightest decile depending on whether the type is positive or reversed. Reversed type, light letters on a dark photo, is where a sloppy sampler returns the background color and the new word vanishes. Store the polarity. The renderer needs it as much as it needs the family name.

Warped and logo typefaces are the honest limit of a library match. A match can find a near weight and a near width. It cannot become a custom wordmark. When the style stage’s weak-guess flag is set, the product should say so in the UI before the user spends a render. That flag is also what a later human proofread sorts on. You do not want those images in the same “looks done” pile as a clean reprint of a gothic on a white menu.

Render glyphs, do not hope the model spells

The last job paints the new content into the image so it belongs: layout, alignment, perspective, composite. There are two implementations, and they do not fail the same way.

The first asks a generative model to draw the letters. Asking a model to draw a specific string, the way DALL·E renders lettering, rarely produces clean, readable glyphs. You get extra strokes, fused characters, and spelling that drifted because the model does not have a character grid. It has a prior over pictures that contain something letter-shaped. That path can be useful for a texture or a sign that only has to look like language from across a room. It is the wrong path when a person will read a price.

The second path is the one that behaves like a word processor. Recognition produced a string. The user edits the string. A real text rasterizer draws glyphs from an actual font. The composite blends those glyphs through the warp you measured. In a purpose-built editor, clicking a region and typing a replacement makes the backend erase the old ink and imitate those letterforms with the new content. Imitation can mean “pick the nearest library font and match color,” which is honest, or “synthesize outlines,” which is research. Either way the letters are rendered, not dreamed. Spelling then fails only if the target string was wrong, which is a data bug you can test, not a sampling bug you cannot.

Composite order matters. Fill the hole first, on the original plate, then draw glyphs on top. If you draw glyphs and fill in one joint sample, you have gone back to the first path and you will not be able to correct a single character without resampling the whole patch. Joint sampling is why a one-word change alters the background noise and the neighboring letter you did not mean to touch. Keep the operations separable so a proofread can rerun the renderer without paying for another fill.

Perspective is a rasterizer feature, not a prompt adverb. If the source quad is a trapezoid, the glyph run has to be transformed into that quad after rasterization, or the rasterizer has to draw in that space directly. A rectangular overlay on a receding storefront is the sticker look, and no amount of color match hides it. Shadows and strokes belong in the same transform. A shadow that falls in screen space while the word falls in sign space is another sticker cue.

If you want one browser component to run every stage on a finished raster, that is the job people mean when they ask a tool to edit text in image. Judge that component as a pipeline with a confirmable string and a mask, not as a magic button. If it cannot show you the string it read, you cannot debug it. If it cannot leave the original file untouched, you should not run it on the only copy.

The render stage is also where critical information goes to die if you picked the generative drawer. A discount, a date, a phone number, a dosage: these are strings with a checksum in the reader’s head. A pretty wrong string is worse than a visible failure. Prefer a rasterizer that cannot misspell over a model that usually spells. When the style match is weak and you still choose the generative drawer, put a human proofread in the path and do not auto-publish.

The harness behind the comparison

What follows is a method, not a table of rates. The quantitative part of this project is a literature review plus a reproducible comparison plan. I leaned on official documentation and product notes, on academic papers, and on third-party reviews. Docs, papers, and reviews are restricted to the 2020–2026 source window, so this review year is the end of that window and the notes are not stale. The one-shot style prototype from the style stage is in that literature. So are the vendor descriptions of how each product wants to be driven.

The sample plan is 3 categories, about 20 images each, 60 total.

The categories are the ones that change which stage fails:

Simple printed text. Black type on white, or any clean ink on a quiet ground. This is the distribution demos already live on. If a tool cannot do this, it cannot do anything harder.
Text on complex backgrounds. Street signs, billboards, words with busy artwork or photography behind them. This is the inpaint stress test, and it is also a detection stress test when the artwork has its own edges.
Special-style text. Handwriting, or commercial logo typefaces. This is the style-stage stress test. A library match can look confident and still be the wrong brand.

Each image uses moderately sized target lettering, not tiny lettering, so detection starts fair. I did not want a failure that was really “the stems were barely a few pixels tall and the task was unreadable to a person.” Tiny type is a separate experiment. Mixing it into this set would let a detector problem masquerade as a renderer problem.

The edit itself is fixed per image. Either replace the selected lettering with a new phrase, or delete the lettering. Use each tool’s intended method: recognition and a click, a direct replacement, or a prompt. Do not force a prompt-only tool through a click-to-edit UI it does not have, and do not refuse to use the auto boxes on a tool that spent its engineering budget on them. You are testing the product’s path, not your ability to misuse it.

These metrics were planned. This article does not report results for any of them.

Recognition accuracy on the result. Run a standard reader on the edited region and check that the new string came out. From those checks, compute a correct-replacement rate. The rate is a method. I am not publishing one.
Visual consistency. Human raters would score each edit from 1–5 on how well the new lettering blends in tone, lighting, and sharpness. An image-similarity metric such as SSIM could compare the original background with the repaired background. No average of that ordinal score appears here, and no similarity figure appears here. If you compute similarity, compute it on the repaired region plus a ring of context. A full-frame score stays flattering because most pixels never changed, which hides a bad local repair.
Subjective aesthetics. Reviewers compare before and after and judge naturalness and visible flaws. An A/B blind pass is optional, not something I am claiming to have tallied.
Automatic quality around the scar. Where you can, quantify the repair footprint with color balance or texture consistency in a band around the edit. Also planned. Also not a number in this piece.

The statistics plan matches the metrics. Per tool, compute the average replacement success and the average visual score, plus a variance so a single lucky image cannot define the tool. Where the sample allows, a t-test or a non-parametric test would check whether a difference between camps, or between a dedicated editor and a general scene fill, is more than noise. This article has no test statistic and no p-value. Without measured data, the expectation is stated in words. Vendor claims and industry feedback point the same way: font matching in the dedicated camp is expected to land high, and a general scene fill is expected to drift when the background is complex. Read that as an expectation, not as a rate I measured and then rounded away.

Reproducible steps are part of the method, or the comparison is a vibe. Log every tool’s real workflow. For an OCR editor: upload, automatic detection, click the region, type the replacement, download. For a desktop fill: record the selection geometry and the fill parameters. For a chat editor: save the prompt and the steps that indicated the region. Capture before and after for every scenario. Have more than one rater repeat the scoring so one person’s taste is not the metric. Hold the environment as still as you can: similar network conditions, the same image resolution across tools, the same target phrase on a given source. With more resources I would add type style and multiple languages as factors, and I would measure editing time and the feel of the UI. This write-up does not report those times.

A note on leakage, because this is where image comparisons usually lie to themselves. Do not tune the target phrase after you have seen which tool fails it. Do not drop an image from the set because one tool looked bad, unless you drop it for every tool and write down why. Do not score a tool on a screenshot of a screenshot if the other tools got the original file. The harness is boring on purpose. Boring is what makes a later number mean something. I am not inventing that later number in this text.

The same comparison, told as a full pass over the products rather than as a stage contract, is the piece on whether current models can edit text inside an image well enough to trust. This page stays with the pipeline, the harness, and the matrix.

The matrix, row by row

The dimensions that matter when you integrate one of these products, or when you refuse to, are narrower than a feature grid. What does it actually edit: a string, a region of pixels, or a background with no concept of glyphs? What is it good at when the stage matches the product? Where does it structurally give up? What is the price shape, without pretending a price is a quality score? Where do the bytes go? Who is the best caller?

Read the rows in that order. A dedicated-camp fit means a font-matching editor. A general-camp fit means a scene tool whose lettering is the unreliable stage. A removal-camp fit means delete only. One row never leaves the device and is not a lettering editor. One row only cuts out a background. One row is an unreleased demo for garbled glyphs in generated pictures. The pattern, once you sort that way, is stable. Some products are creative processors with no dedicated lettering editor. Some are friendly brush or prompt entry points. One family only deletes. The purpose-built lettering editors are the rows that prioritize font matching and automation.

Tool	What it edits	Strengths	Limits	Price	Privacy	Best fit
Photoshop	Delete or replace selected elements, including lettering, and generate image content to fill the erased area. The fill step is Generative Fill.	Powerful scene context, fine control over selections and parameters, and it works with layer effects. The generative model family here is Firefly.	No dedicated text recognition or font matching. Generated lettering is often generic and not guaranteed legible. Steep learning curve, and the work is manual.	Subscription.	Edits upload to the Adobe cloud. Login is required. Copyright rules apply.	General camp, for professional scene work: complex scenes, object replacement, background extension. Not for precise lettering replacement; that still needs a manual font match on a real type layer.
Canva Express	Brush-select plus a text prompt to replace a selected element, for example deleting or replacing lettering. Magic Erase removes objects or lettering. Magic Edit is that prompt brush.	Simple and browser-based, with no install. It integrates with templates. Element replacement is fast on ordinary images.	Still in beta, with limited precision. Weak on complex backgrounds and fine detail. No auto-OCR, so you mark the region yourself.	Basically free. Advanced features need a paid Pro plan. A chat plugin can call it without a separate charge.	Online processing, and an account is required. The privacy policy stresses asset safety, but processing is still in the cloud. Output is subject to copyright and community rules.	General camp. Content creators doing quick element edits, or deleting and replacing lettering on flat design and social graphics, where pixel-perfect fidelity is not required.
ChatGPT + image editing	Upload an image and describe edits in natural language, including adding or removing lettering. In plugin mode it can call a desktop editor or a design app, for example blurring a background or replacing lettering through those plugins.	Flexible and iterative, with multi-turn consistency. The Images 2.5 model focuses on precise editing. Templates such as Sketch, plus prompt sharing, are part of the workflow.	Limited recognition and rendering. Expect errors or garbled spelling when lettering is drawn by the model. No exact font match.	Free, and Plus, Pro, and Enterprise tiers. Image editing is a general capability. Desktop plugins are reachable through the chat.	OpenAI is privacy-respecting in its framing, but uploads may be used for training, with an opt-out. Plugins require login and follow their own rules.	General camp. Creative prototyping and fast iteration: natural-language edits, quick slogans, color changes. Not for fine typographic proofing.
Pixelmator Pro	General editing and repair, including machine-learning content fill and object removal, plus built-in enhancement and canvas tools.	A friendly interface, Apple machine-learning features, and a smooth feel. Multi-layer and vector tools for creative work.	No dedicated text recognition or replacement. Mac and iPad only.	One-time purchase or a Mac App Store subscription. Cheaper than a full professional desktop subscription.	Mostly on-device. Whether anything is uploaded depends on what you do. No extra privacy exposure from the editor itself.	Routine editing for Mac designers: object removal and color adjustment. Not a lettering editor.
Remove.bg	Automatic background removal only. No lettering edit.	Focused and efficient. Fast, accurate deep-learning segmentation.	Single-purpose, and unrelated to lettering edits.	Basic free. Paid plans cover batch and high-resolution removal.	Web or API use requires an upload. The policy says content is not stored, or is kept only briefly in order to process.	Background removal for commerce and cut-outs. Useful as preprocessing before manual lettering work. Not a lettering editor.
Runway ML	Cloud image and video generation and editing, including object removal, expansion, and video inpainting.	A creation platform for stills and motion, with collaboration and an API. Frontier models include Gen-4.5 and GWM.	Aimed at creative generation and video, not dedicated lettering-object editing. Some advanced features need paid credits.	Free basic tier, pay-as-you-go, plus team and enterprise plans.	A cloud service that caches processed content. Check the terms on using assets for product improvement and on commercial licensing.	Creative video and interactive work. It can delete lettering as a side effect. It is not a font match.
ClipDrop	Cleanup removes a selected object, person, or lettering. Text Remover deletes lettering and, as the product describes it, rebuilds what was behind the text.	Simple, generally natural results, including on complex backgrounds, plus a free allowance.	Can only delete. It cannot replace or add lettering, cannot keep the original style, and cannot re-typeset. Needs a network.	Free daily quota. Premium is a subscription.	Account login for the web app. Images are uploaded to the cloud under Stability AI policies.	Removal camp. Clearing unwanted lettering, watermarks, or blemishes. Replacement is a different tool’s job.
Fotor	Auto-OCR on text regions. Click a detected region and type the replacement. The backend removes the old lettering and imitates the same letterforms with the new content. Prompt-based edits are also available.	Purpose-built for lettering in images. Auto-matches font, color, and effects. The flow is close to a word processor. It can delete, replace, and add lettering.	Needs a network, and server time varies. Weak on tiny type, handwriting, and warped type.	Free tier can watermark. Premium unlocks batch and high-res.	Images are transferred encrypted. The policy claims no retention for online editing.	Dedicated camp. Content marketing, ads, promo pages, social images, and typo fixes where font fidelity matters.
ReWords AI	Detects existing lettering, removes it, rebuilds the background, and renders a translation or replacement with matching font, color, and size. Built for finished images when there is no source file.	Aimed at flyers, posters, menus, and labels in the browser. The everyday job is localization and correction without rebuilding the layout.	Results depend on image quality, font complexity, and background texture. Complex photographic backgrounds are best-effort. Not for IDs, invoices, or official documents.	Free first edit after sign-in. Credit packs and subscriptions cover volume and higher-resolution exports.	Cloud processing. Read the privacy terms before a sensitive upload.	Dedicated camp. You do not have the design file, and you need the wording changed rather than the scene reinvented.
Textify (Storia.ai)	Mainly fixes garbled lettering inside generated images, replacing wrong glyphs with a meaningful string.	Designed for generated-art workflows. It improves readability somewhat and offers a one-click replacement.	Only pseudo-text in generated images, not real photos. Current results are limited, and examples still show spelling errors.	Not officially released. It is a project demo.	No public privacy policy. The project is still in development and testing.	Hobbyists repairing garbled lettering in generated pictures such as Stable Diffusion. Immature, and not for real photos.
Baidu Netdisk AI retouching	Adds image-text recognition and editing inside the drive app. Select a region and change font, color, and size.	Short steps. Auto-detects regions and offers edit controls. Advertises editing in the same font format. Saving and sharing stay inside the same ecosystem.	The app is required. Accuracy depends on scene complexity. Output quality is thinly verified in public.	Built into the drive app. Access may depend on a VIP tier. Described as currently free.	An internal feature. Processing may happen on the vendor’s servers under the drive’s general privacy rules, including encrypted storage.	Dedicated camp. Quick replacement for people already in that ecosystem, such as ads and layout notes. Watch the output.

That finished-image row — detect, remove, rebuild, render, in a browser, with no design file — is the one you open when the export is all you have. The matrix is the integration view. The passes themselves were less tidy than any single row.

What the passes showed, without a scoreboard

Automatic lettering edits inside images are not a mature, single-purpose capability, and coverage is still split by what each product actually automated.

The dedicated camp is the one that preserves font style and turns a replacement around quickly. That matches the architecture: a box, a string, a fill, a rasterizer that imitates letterforms instead of sampling them. The general camp wins on image consistency and on creative flexibility, and it lacks fine control of type. A scene fill can remove lettering and rebuild the background without an obvious seam, and the letters it invents are still not style-consistent. A natural-language editor can add or change wording in one pass, and it still returns typos or warped glyphs often enough that you cannot skip a proofread. Those are qualitative splits. They are not recognition rates, not similarity scores, and not significance tests.

One observation held across the set, and it maps directly onto the inpaint stage. Lettering on complex backgrounds — texture, lighting gradients — failed or left artifacts far more often. Printed lettering on a quiet ground came out well in almost all of the tools. If your image is in the easy category, you are in the easy zone, and you should not brag about it as if it were the hard zone. If busy photography sits behind the words, expect a fight, and budget a human look at the hole before you accept the composite.

I am not going to dress that observation up as a computed footprint. The color-balance and texture checks in the harness are how I would quantify “artifact” on the next pass. They were not run to a published figure for this piece. The human ordinal score was not averaged. What I will stand behind is the direction: quiet grounds are kind to every camp, and textured grounds expose the fill. The style stage has its own direction, separate from the fill. Logo typefaces and handwriting are where imitation stops being imitation and becomes a different word in a nearby font. If the brand lives in the outline, do the change on a real type layer or do not do it.

Spelling failures deserve a developer’s reading, not a shrug. When a rasterizer misspells, the target string was wrong or the font did not contain the character, and you can assert on that in a test. When a generative drawer misspells, the sample drifted, and the next sample may drift differently. The second bug is not something you patch with a prompt adjective. You route critical strings away from that drawer. The passes made the routing obvious even without a rate beside it.

The short answer the stack forces

Shipped products fall into three camps, and each camp wins on a different axis.

Dedicated editors lead on font fidelity and speed. They are built to keep the original style, to box the words, and to let you type. That is the path that respects the stage contract: a string you can see, a mask, a fill, a rasterizer.

General generative tools lead on scene context and on creative changes to the picture as a picture. They understand more of the surrounding scene than a font matcher does. Their generated lettering is often blurry, misaligned, or misspelled, because lettering is the stage they refuse to isolate. Use them when the job is the scene. Do not use them as a type engine.

Removal-only tools delete lettering cleanly and cannot put words back. They are a fill stage with a UI, not an editor. That is useful, and it is not the feature the headline promised.

No single tool does everything well. The headline feature everyone wants — type new text and have it look native — is still the weakest link. In the passes I cared about, clean removal was easy. Convincing replacement was not. The stack explains why. Removal can stop after a decent fill. Replacement has to survive detection, style, and a renderer, and readers notice a bad glyph faster than they notice a slightly wrong brick.

Do not read the camp split as “only the dedicated camp is allowed.” Scene reconstruction, object removal, and extending a background are real jobs, and the general camp is built for them. The mistake is asking a scene model to also be a type setter, or asking a type setter to invent a new background object. Pick the camp that owns the stage you are actually calling. If you need two stages from two camps, run them as two calls and keep the intermediate, so you can throw away the bad call without rerunning the good one.

Who should run which camp

If you are a content creator, the usual job is small and specific. A discount changes. A date changes. A label gets a translated line. Prefer a dedicated editor for that. The flow feels like a word processor: select the words, type, let the backend rebuild the ground and keep the style. The barrier is low because you are not asked to become a scene compositor. For a price, a date, or a translated line on an export, edit text in image on the finished file instead of reconstructing the layout from scratch. For heavier creative changes — overall color, background objects, a different composition — add a general scene edit, and plan to touch up the output. Avoid relying on directly generated in-image lettering to carry critical information. The error mode is wrong characters that still look intentional. That is worse than an obvious hole.

If you are a designer, the bar is precise visual consistency, and a font match you do not control is not precise. Use a scene fill as an aid: remove the old lettering with it, then set the new wording yourself on a type layer so the family, the kerning, and the brand rules are yours. A chat product’s desktop-editor extension can streamline part of that flow, especially when you are still deciding what the new line should say. A browser brush-and-prompt edit can speed prototypes and internal reviews. Check the pixels before they leave the review. Treat automatic replacement as an assistant, not as a one-stop type department. The assistant is allowed to be wrong. Your file is not.

If you are on a legal or compliance team, the risk is privacy and copyright, and it is a data-flow risk rather than a style opinion. Every cloud service in this set — chat, browser design tools, dedicated web editors — requires the image to be uploaded. Upload implies a leak surface and, on some of those services, a training-use question. Read the vendor’s privacy terms. Confirm that retention, region, and authorization match the policy you already apply to customer artwork. An opt-out that exists only in a personal account setting is not an enterprise control until you have tested that it binds the workspace you actually use. Output raises a second issue: a rewritten logo outline can step on the original creator’s rights, and a generated sentence can state a fact you did not approve, including a price or a claim. For official documents and for ID photos, do not use an automated tool to edit the lettering. The downside is a dispute, not an ugly kerning pair.

On-device editing is the exception that proves the rule. If the pixels never leave the machine, the training-use question changes shape. The moment a “local” app calls a cloud fill for the hard stage, you are back in the upload regime for that stage. Ask which stage is local. Do not accept the word local for the whole pipeline because the chrome is a desktop app.

The safest approach is a mixed toolkit, with the original file kept untouched. For a simple string change, start with recognition plus a dedicated editor. For a creative edit, or when the source file is missing and the scene itself has to change, use a desktop fill or a chat model as a first pass, then proofread. For clearing unwanted lettering, use a removal tool, and do not expect it to type. Stay critical of the output. Compliance teams should write a policy for image editing and for the review of generated content, including who is allowed to upload customer files and which document classes are out of scope no matter how good the demo looked.

A practical policy is short enough to fit on one page. Keep the source object immutable. Store the edited object beside it, not instead of it. Require a human proofread when the string is a price, a date, a name, a legal line, or anything regulated. Forbid automated edits on identity documents and on official records. Record which vendor processed the file, because the privacy review is per vendor, not per category. None of that requires a new model. It requires the sidecar the stack section already asked for: operation, target string, and whether a person confirmed the read.

What I would actually ship

The models can change pixels that used to be letters. Whether they can do it well is a separate question, and the answer still depends on which stage you needed. Dedicated editors, built to detect, erase, match, and rasterize, are the ones that take the headline feature seriously. General-purpose generators still cannot reliably draw lettering that belongs in a design. Removal tools remain good at the job their name implies, which is the fill, and only the fill.

If I were reviewing a pull request that added “edit the text in the picture” to a product, I would block it unless the interface showed the detected string, the mask, and a way to edit the string before the fill ran. I would block a path that sent a price or a legal line through a generative drawer. I would block any path that overwrote the only copy. I would ask which camp owns each stage, and I would not accept one vendor checkbox as an answer for all of the stages. I would also ask what the harness is. A single hero screenshot is not a harness. The category split in the method section is the minimum: quiet type, type on a hard ground, and type whose style is the brand. If those situations were not tried, the feature is a demo.

Ship the mixed toolkit. Keep the original. Proofread the glyphs at actual pixel size, not in a thumbnail. Write the policy before someone pastes an identity document into a cloud fill because the button was convenient. The stack is understandable. The weakness is specific. It sits in the last mile, where new letters have to look as if they had been printed with the old ones, on a surface the fill had to imagine. That mile is still the hard part, and any roadmap that marks it done because a clean word on a white card looked fine has tested the easy category and called it the product.

Detection Boxes to Redrawn Glyphs: A Developer Pipeline for Text Inside Images

baiwei — Fri, 09 Oct 2026 08:43:05 +0000

Ten ways to change the words baked into a finished image — and keep the layout, type, and background a designer already signed off.

You do not translate an image by translating a string. You find boxes, read what is inside them, and put different glyphs back into the same pixels. A campaign poster leaves design as English. A menu leaves as Japanese. A product label leaves as German. A diagram leaves with Spanish callouts. What you hold is the export: PNG or JPEG, layers gone, type painted into the raster. The brief does not change. Ship the other language. Do not ship a different design.

A document translator can let the layout flex. An image cannot. Column width, drop shadow, and the texture under a headline are pixel counts. Erase the source word and you also erase the pixels it covered. Image localization is that constraint: meaning has to change, and the picture is not allowed to.

This piece follows the pipeline a developer would build, then the ten real ways to own each stage. Scripts, OCR, and APIs come before any phone-camera story. Fidelity, cost, and failure modes stay attached to the method that has them.

The Detection Box Is the Unit of Work

Treat every text run as a rectangle, then as a string, then as a hole you have to fill, then as ink you have to put back. If you build the loop in Python, the stages are boring on purpose. Boring is what you can test.

Detect. Find the regions that contain glyphs. OpenCV edge or contour detection will give you boxes on high-contrast labels. A deep OCR model — EasyOCR, or Tesseract in a mode that returns word boxes — does better on photos. You want coordinates, not one blob of page text. A box that is too tight clips diacritics. A box that is too loose eats a logo or an arrowhead.
Read. OCR the crop. Language hints matter. Spanish on a label is not the same pass as Chinese on a package. Keep the confidence score. A low-confidence box is a box you should not silently translate.
Translate. Send the string, not the image, to a machine-translation API or to a person. Google Translate, DeepL, and Baidu Translate are the usual engines here. Lock tokens that must survive unchanged: brand names, SKUs, registered marks, and numbers whose meaning is not linguistic.
Erase. Cut the glyphs out and reconstruct the background. On a flat fill this is a rectangle of the sampled color. On a photo, gradient, or texture you need an inpaint. A small script usually fills with a flat color or a neighborhood average. A diffusion inpainter is what you call when that average looks like a bandage.
Redraw. Draw the translation into the box with Pillow or OpenCV. You choose the closest font you actually have a file for. You set weight, fill, and size. You measure the advance width. If the string does not fit, you shrink, wrap, or track. You do not pretend it fits.
Write. Save a bitmap at the same resolution you loaded. Keep the original. Diff them. The diff should light up inside the old text boxes and almost nowhere else.

That is the whole product, whether a person runs it by hand or a job runs it overnight. Hosted tools hide steps 1 through 5 behind one request. Multimodal models hide them inside a prompt. Diffusion models ask you for the mask and then invent the pixels. A designer does the erase and the redraw with a stylus and a font menu. The phone in your pocket does a coarse version of the loop and stamps a reading overlay on top, which is a different product.

Atlas Cloud has described the same split: after OCR lifts the words, someone still has to place them back into the design. A pure raster edit merges the change onto a new layer and gives you less control over individual letters. Manual typesetting is slower and, when the font is right, the most precise. A script sits between those two. You get coordinates. You do not get the original text frame, the original leading, or the original OpenType features unless you reconstruct them.

Input is a bitmap. Output is a bitmap of the same size. The value every method either computes or fakes is a list of boxes with strings. If you cannot see those boxes, you cannot debug a single bad caption.

Four Constraints a Redraw Has to Survive

The layout is frozen. The words are not. Every method below is a trade against the same four limits.

Length is not preserved. A short English phrase often grows by 30% or more in German or French, and usually shrinks in Chinese or Japanese. The box does not grow with the grammar. You resize, re-wrap, or re-track. Assuming the translation fits unchanged is how buttons overflow.

Type has to be rebuilt. Family, weight, color, letter-spacing, and any shadow, stroke, or outline are part of the design. A right sentence in a default sans still reads as a patch. Without the identical font file, a full match is hard. Photoshop and Pillow share that limit: they can only render a face you can load.

The background under the glyphs is destroyed when you erase them. Solid color is easy. A photo, gradient, or texture is reconstruction. Feather the edge and check the seam at 1:1. A visible rectangle means the fill failed.

Non-text geometry is in scope. Arrows, callouts, numbering, and legends contain text or point at it. OCR returns strings. It does not return "this arrow belongs to this caption." A person, or a model that sees the whole figure, has to.

The right method depends on fidelity, how much text is in the frame, and whether the source file still exists.

Scripts Before Buttons: a Custom Pipeline and OCR Typesetting

Own the batch job before you rent a button. The phone-camera path is last on purpose.

Method 8: A computer-vision script you run yourself

How it works. This is the loop from the first section, as code. Detect regions with OpenCV edge or contour detection, or with a deep OCR model. OCR and translate each box. Cut the glyphs out and fill the background. Draw the new string. Write the file. There is no off-the-shelf GUI. The script is the product.

Tools. Python with Pillow, OpenCV, EasyOCR, and pytesseract. A Baidu or Google Translate API for the string. The same job exists in Adobe PixelSense and in C# or .NET image code.

Input and output. A normal bitmap in. A same-resolution bitmap out. Keep the box list beside the file so a bad label can be blamed on OCR, translation, or the draw.

Automation. High once the script is actually finished, and fully batchable. Thresholds, language packs, and font fallbacks are the unfinished part.

Fidelity. Medium, and it depends on those details. You can roughly honor layout by drawing into the original boxes. You usually cannot match fonts automatically — you pick the closest face by hand. The fill is commonly a flat color or a neighborhood average. That is less natural than a diffusion inpaint, and it fails on complex backgrounds and non-standard lettering.

Time and cost. Development cost is high. Runtime is on the order of seconds per image. Money can sit near zero next to a third-party service if the pieces are open source or already licensed. The invoice is the coding.

Pros and cons. Full control, batch, and a path that can stay off the public network. Design-detail fidelity trails a dedicated image model. You will tune parameters and correct misses.

Best for. An enterprise process requirement, or small-scale label replacement. The usual case is a merchant scripting OCR, translation, and write-back on product-image labels. Demanding, and flexible.

Method 2: OCR, then typeset like you still have a layout

How it works. Same front half as Method 8 — OCR for content and position — then a person or a layout app owns the back half. Translate by machine or by a translator. Re-render at the original positions and styles. The pipeline is recognition, translation, layout reconstruction, output image.

Tools. Tesseract, EasyOCR, and Adobe Acrobat OCR. Google Translate, DeepL, and Baidu Translate. Layout in Adobe InDesign or Photoshop, or in Pillow and Matplotlib when the figure is a chart.

Input and output. PNG or JPEG in, a new image out. A PDF or AI detour is worth taking: a text frame can change leading without you repainting a gradient.

Automation. Medium. OCR and translation script cleanly. Final typesetting usually does not, once source and target lengths diverge and someone must change size and line spacing without hitting a photo.

Fidelity. High when fonts and frames match. Expansion still breaks the original spacing, so you scale or re-wrap. A bad box on a busy figure hurts more than a weak synonym. Kerning and ligatures drift even in the right family. Leader lines and numbering are not handled. They stay manual.

Time and cost. OCR plus translation is seconds to minutes. Typesetting is minutes to tens of minutes. Money stays low on open-source OCR and an API you already pay for. The labor is the typesetting hour.

Pros and cons. Faster than a blank-layer redraw, and reviewable while the text is still text. You depend on OCR accuracy and font matching. One button will not publish this.

Best for. Chart annotations, technical manuals, and educational diagrams. PDF translators follow the same idea: parse structure, translate region by region, replace in place, preserve the layout. On a flat image you are running that algorithm without a text layer.

Call an API When You Want the Model to Repaint

Delegating detection or inpainting does not delegate the check on numbers and names.

Method 6: Multimodal models that take an image and an instruction

How it works. GPT-4V, a later vision-capable GPT-4-class model, ChatGPT image editing, or a DALL·E edit takes the image plus an instruction to translate the text and leave the design alone. The model is supposed to detect, understand, and generate together. What you can operate is: image and instruction in, recognition and translation, a new image, then a human proofread. Do not treat a follow-up question as a box list.

Tools. OpenAI ChatGPT with image features such as GPT-4V. The DALL·E 3 image-editing API when you call it. Google Imagen Editor, mostly research-stage rather than a production control. Inside a service, developer access is the OpenAI API.

Input and output. Image plus prompt in. A new image out, at the original resolution or a size you set. You do not get layers back. If the model regenerates the frame, inspect more than the caption.

Automation. High. One prompt can carry the job. Log the prompt, the model, and the image hash anyway. A prompt is a bad audit log.

Fidelity. Generally high on layout and style. OpenAI's examples include a coffee-machine instruction diagram translated into Spanish with the layout kept. Atlas Cloud has reported GPT-4V translating infographic text and leaving the rest of the figure intact. Models still miss strings, misspell them, and shift elements when the new length is far from the old one. Budget a fix. This is not fine typesetting. It is often good enough for marketing art, which is a different bar from a regulated label.

Time and cost. Medium. One image is often seconds to tens of seconds, plus an API or subscription cost and your review time. Quota is finite.

Pros and cons. Arrows, legends, and hierarchy can survive, because the model sees the figure rather than a bag of boxes. Fine control is limited. Expect spelling errors, leftover source language, or compositing artifacts.

Best for. Infographics, posters, and illustrated promotions. A usable prompt is concrete: translate all text into Chinese, and do not change anything else. When one line overflows, redo that region. Redoing the poster is how the background drifts.

Method 7: Diffusion inpainting, where you own the mask

How it works. Mask the text, then prompt a replacement that keeps the background. DALL·E 3 or Stable Diffusion inpainting fills the mask. The flow is detect, build a mask, prompt ("replace X with Y, keep the background unchanged"), fill, inspect. Uploading the original and a mask to an editing API is the same idea over HTTP.

Tools. The OpenAI DALL·E 3 editing API. Adobe Firefly Generative Fill. HuggingFace Diffusers, including runwayml/stable-diffusion-inpainting, when you run the weights. ControlNet on text detection, and LaMa for the inpaint, when the mask should follow glyphs instead of a hand-drawn blob.

Input and output. Original, text mask, and prompt in. An image out. The model is supposed to take language and style from the prompt. Check the glyphs. These models are not spellcheckers.

Automation. Medium-high. Generation is one call. The mask is not. A person draws it, or OpenCV or OCR emits it. Too small and old serifs remain. Too large and the model repaints a product that was fine.

Fidelity. High, a step under models built to edit text. Long strings break words. Style tracks the prompt. The strength is background repair: a fill can blend into a complex texture more naturally than a neighborhood average. The weakness is letterforms — slight blur, a bad shape, an extra detail. Tuned masks and prompts are the difference.

Time and cost. Medium. Typically ten to tens of seconds, or your local GPU. Cloud adds queue time. You pay in compute or an API fee.

Best for. Creative work where a slight style shift is acceptable: concept art, slogans in illustrations. The banner fixture is Stable Diffusion taking "Summer sale — 30% off everything" to "Sommer-Sale — 30 % auf alles," background held, lettering re-rendered. Short lines survive more often than paragraphs. Set a regulated number with a real font.

Method 3: Format-preserving document services

How it works. Document translators parse layout and fonts, machine-translate each block, and re-typeset. Steps: layout parsing (coordinates and fonts), translation by a cloud service or a local engine, re-rendering, file write. Graphic images are typically left untouched, which is right for a scanned brochure and wrong for a figure whose labels had to change.

Tools. PDFTranslator, routing through Gemini and ChatGPT engines, the Pangeanic ECO platform, Pairaphrase, and Smartcat. Many accept a PDF or an image and return a translated PDF or graphic in the original styling. Integrate them as a service unless you also have an API you are allowed to call.

Input and output. Scans, images, or PDFs — JPG, PNG, or PDF — in. A file in the same general format out, with layout, tables, and graphic positions essentially where they started. A table that reflows by a row is still a defect.

Automation. High, end to end, with a light human proofread and a heavier one where numbers live. Unattended publishing is a different claim.

Fidelity. High. Careful OCR plus a dedicated translation system or an LLM, aimed at coordinate-level mapping. Layout and fonts are preserved. Document-level fidelity in this class is described as reaching 90% or better. Remaining errors are mostly length, such as a German cell in an English-sized column. Unusual layouts and handwriting are weak. A strong financial report does not imply a strong whiteboard photo.

Time and cost. Seconds to minutes for most jobs. A 60-page PDF is a matter of minutes. Paid service, or purchased API quota. Bulk and long documents are where the bill shows up.

Pros and cons. Automation and fidelity, against fees. Weird layouts and handwriting are poorly supported. A flat PDF back means the next fix looks like Method 2.

Best for. Enterprise technical documents, product manuals, training material, and financial reports that must keep their layout. Under the hood it is still OCR, translation, and automatic typesetting. Debug it like a pipeline.

A Hosted Loop When You Do Not Want to Own the Mask

When the source file is gone and the image is ordinary marketing art, a tool built for this loop is the short path that still tries to keep the design.

Method 4: Online AI that detects, translates, and repaints

How it works. Upload. Auto-OCR the regions. Translate, with a chance to proofread. The model erases the source glyphs and repaints the translation in a similar face. Download. OCR, machine translation, and image completion are one flow.

Tools. ReWords AI, ImageGPT, EditTextImage, Codia, and VisualGPT. Useful interfaces let you edit a suggested line and lock a line that must not change. A lock is how a brand name survives.

Input and output. PNG or JPEG in. A same-resolution bitmap out, translation rendered close to the original style. Not a layered source file. Layers mean Method 10 or Method 1.

Automation. High for one asset: upload and generate. Batch becomes credits, an API, or a queue.

Fidelity. High overall. Fills often match font style and background texture well enough that the edit is not obvious. Dense text, complex lighting, and decorative fonts still flaw. OCR tracks source quality. Neural translation varies and needs a human review. A clean look is a draft, not a proof.

Time and cost. Low to medium. Seconds to a minute. Nothing to install. A free allowance is normal. ReWords includes a few free generations per image, then paid credits. Bulk means payment or credits.

Pros and cons. You replace text on the image without the design file. Some images need a retry or a tighter instruction. Copyright and sensitive content stay yours.

Best for. E-commerce images, ads, social graphics, menus, flyers, and comics. On packaging, "SUMMER SALE" becomes "夏季特卖" with the background unchanged. An EditTextImage-style line turns "Summer sale — 30% off everything" into "Sommer-Sale — 30 % auf alles" with the layout left in place. A dense plate with leader lines belongs back in Method 2 or Method 3.

On a service built for this, you translate text in an image with a short runbook. Upload and choose the image-text function — a "SUMMER SALE" banner is the fixture. Pick the language, correct the line-by-line suggestions, and lock brand names. Generate or apply: the original glyphs come out, and the translation is redrawn in the original color, a similar face, and the existing background. Download and review at 1:1. That is the default when there is no source file and the clock is short. Numbers and proper nouns still fail, for the same reason they fail in a script.

Source Files, a Designer, a Studio, and a Phone

These four are what you do when you still hold the design, when fidelity outranks speed, or when you only need to read a sign. Two of them are the highest-fidelity options. One is the cleanest path if the source exists. One is never the file you publish.

Method 10: Edit the source, not the export

How it works. If the bitmap came from vector or layout software, replace the string in the text frame. InDesign or Illustrator, with scripts or plugins, can pull text out by OCR or by conversion, translate it, and put it back. Another route is PDF, then OCR and translate, then import. Effects and layers already exist. You are not guessing them from a JPEG.

Tools. Adobe Acrobat for OCR and an export to Word. Illustrator find-and-replace. InDesign scripts. Lokalise, or Transifex's InDesign plugin, when the strings belong in a translation memory.

Input and output. An editable document or a PDF in. Text replaced in the original file out. The PNG is a build product.

Automation. Medium, and only as scriptable as the design app. Swapping strings in tagged frames is automation. Guessing which outlined text used to be which sentence is not.

Fidelity. Very high, because you edit the source and layers survive — if you actually have the file. A flattened export is Method 2, whatever someone named it.

Time and cost. Medium. Replacement still needs a person for overflow. Cost stays low when the licenses already exist.

Pros and cons. Precise typesetting from the real file. No file, no method.

Best for. Work where you hold the design source, such as an InDesign manual. The emergency variant is a rough OCR pass for reviewers, then a refine inside the frames before anyone prints.

Method 1: A designer redraws it

How it works. A person runs the pipeline. Extract the text by OCR or by typing it off. Translate and proofread. Find or recreate the face. In Photoshop, Illustrator, or a similar tool, erase, set the translation, and rebuild effects. Order: extract, translate and proof, match the font, erase, insert, adjust.

Tools. Adobe Photoshop, Illustrator, InDesign, plus GIMP and Figma. Tesseract can speed extraction. It should not speed approval.

Input and output. Any image in, including PNG or JPEG. A new image out, or better a layered PSD or AI file. A flatten-only delivery throws away the next edit.

Automation. Low. Only OCR reasonably scripts.

Fidelity. Highest from a person on the pixels. The designer controls font, position, and style. Closeness depends on skill. It is slow and sensitive to expansion. Without the identical font, a full match is hard.

Time and cost. High. Minutes to hours, plus a learning curve on the software, so the tool is not free before the labor. Wrong default at volume. Often right for one flagship piece.

Pros and cons. Layout and style can be held for demanding brand work, including arrows and callouts. Slow, expensive, and easy to miss an accent or a shadow left over from the old word.

Best for. Ad posters, packaging, UI prototypes, and print that must stand next to the original. Atlas Cloud's point sits here: after OCR lifts the words, placing them back is still a design act. A pure raster edit merges onto a new layer with less control over individual letters. Manual setting is slower and, in skilled hands, the most precise option.

Method 9: A translator plus a DTP studio

How it works. You hire the pipeline. A translator translates. A desktop-publishing designer replaces the text. Steps: identify and edit the source if it exists (PSD or InDesign), translate, adjust the layout, re-render in the same software. A bitmap-only job is Method 1 with a real translation process. An InDesign package is Method 10 with bilingual review.

Tools. The Adobe suite, InDesign, Illustrator. CAT tools such as Trados or MemoQ keep terminology stable. The CAT tool does not export the poster. The design tool does.

Input and output. Any image or source package in. A high-quality localized image, and usually an editable file, out. Ask for the editable file. A flatten-only delivery is the wrong artifact at this price.

Automation. Low. People intervene at every stage that matters.

Fidelity. Highest, shared with careful manual design and often stricter, because a translator handles cultural nuance and a DTP operator handles format. You still review against the source and the glossary.

Time and cost. Highest of the ten. Translation fees plus design time, usually far above any machine option. Correct for sensitive work. Wrong for a tile that expires on Thursday.

Pros and cons. The quality ceiling, and the bill. Time and money sit above every automated method.

Best for. Brand campaigns, major trade-show pieces, image content in government or legal documents, and publications that need a human proofread. A wrong numeral is an incident. Put this method next to Method 1, not next to a camera app.

Method 5: Phone camera apps, for reading and only for reading

How it works. Google Translate, Baidu Translate's camera, Microsoft Translator, or WeChat scan-translate takes a photo or an import, runs OCR, translates, and overlays the result or saves a snapshot. Capture, recognize, translate, mark the original, show it. Comprehension on the spot. Not a master.

Tools. Those four, and the cousins inside other apps. No project file. No font menu that matters.

Input and output. A phone photo or screenshot in. An overlay or extracted text out. Some apps save an annotated image. That image is a reading aid.

Automation. Fully automatic. You point the camera.

Fidelity. Low, the bottom of this set. Default face, coarse mask, original layout and style not preserved. OCR and translation can sit near professional engines. The render is still not a design file you can publish. People who photograph text with Baidu Translate and save it often get the app's own mark, or a patch, on the frame.

Time and cost. A few seconds, often real time, usually a free app. Right price for "what does this say." Wrong price to treat as production art.

Pros and cons. Fast, no expertise, enough for daily life. They cannot preserve the original design. A logo or watermark may land on the output. Reference reading only.

Best for. Foreign menus, street signs, and manuals while traveling. Not the ad, not the label, not the file you print. A camera-roll screenshot is a hint at best. Rebuild with one of the other nine.

Ten Methods Side by Side

Fidelity here means how closely the translated image matches the original design, not how fluent the sentence is. Phone apps sit at the bottom of that column because they help you read and they add a large mask. Manual design and a human DTP pass sit at the top because a person protects layout and type. Do not read "low cost" as "best." A cheap path that wrecks the file is expensive on the day you reprint.

Method	Design fidelity	OCR / recognition	Multilingual	Technical difficulty	Cost	Speed	Best fit
1. Pro design software (manual)	Very high	N/A (human)	Any	High (needs a designer)	Highest (labor)	Slow (manual)	High-spec brand / marketing assets
2. OCR + typesetting	High	High (structured text)	Many	Medium-high (software/scripts)	Low (open source)	Medium	Technical docs, labeled diagrams, structured charts
3. Document translation service	High	High (professional OCR)	Very many	Low (turnkey)	Medium-high (service fee)	Fast	Bulk PDF/scans, professional documents
4. Online AI tools (ReWords and similar)	High	High (AI)	Many	Low (turnkey)	Low/medium (free allowance + paid)	Fast	E-commerce posters, comics, everyday multi-image localization
5. Phone camera apps	Low	Medium-high (handheld)	Many	Very low (turnkey)	Low (free)	Real-time	Quick reading for travel / daily life
6. Multimodal large-model editing	High	High (model)	Very many	Low (turnkey)	Medium-high (API fee)	Medium	Infographics, ad posters, UI screenshots needing fast reasoning
7. Diffusion models (DALL·E and others)	Medium-high	Medium (needs a mask)	Many	Medium (prompt craft)	Medium (API/compute)	Medium	Creative design, where a style shift is acceptable
8. Custom Python pipeline	Medium	Medium (tool-dependent)	Many	High (programming)	Low (open source)	Flexible (batch)	Automation environments, private or enterprise batch jobs
9. Professional human + DTP	Very high	N/A (human)	Any	Very high (expert team)	Highest (service fee)	Very slow (manual)	High-end publications, strict text and format demands
10. Direct source-file translation	Very high	High (from the source)	Many	Medium (design software)	Low (if the software already exists)	Medium	Projects that still have an editable source

Phone apps are for comprehension and add a large mask, so fidelity is low. Professional design and a human DTP pass preserve layout and type, so they score at the top. Method 8 is medium fidelity: open-source cheap to run, expensive to build. Method 4 is fast, with a free allowance and a paid tier after it. Method 7 is medium-high because long text breaks. Do not read the cheapest row as the best row.

Three Tutorials With the Code in the Fence

The three tutorials are complete enough to run or to reject. The first calls an image edit API with a mask. The second is the hosted runbook. The third is Tesseract, a translation client, and Pillow. Hash comments stay inside the fences. They are notes to the interpreter, not headings.

Tutorial 1: GPT-4 Vision, or a masked image edit

Overview. This is the smallest program that asks a vision-capable OpenAI model, or a DALL·E-style edit endpoint, to change text inside an image. The fixture translates the English "Summer sale" into the Chinese "夏季特卖."

Prep. An OpenAI account and an API key. A sample image that contains the source words. A mask aligned to that image. The filenames below are orig.png and mask.png. The mask covers the "Summer sale" region and as little else as you can manage. Mask polarity differs by API version. Read the current field definition before you burn a batch.

Flow. Load the original. Call the GPT-4V path or the DALL·E editing endpoint. The model recognizes and translates. The model generates. You download and proofread. Skip the last verb and you have a demo, not a pipeline.

Align the files. orig.png is the source. mask.png matches its pixel size. The mask covers the source phrase and not the product, the logo, or a legal line you did not mean to regenerate.
Call the API. The snippet uses the OpenAI Python SDK shape that takes an image, a mask, a prompt, a count, and a size. The prompt tells the model to translate the masked words into Chinese and to leave the rest untouched.

import openai
openai.api_key = "YOUR_API_KEY"

# Ask the model to translate the text in the masked area into Chinese
# while leaving everything else unchanged.
response = openai.Image.create_edit(
    image=open("orig.png", "rb"),
    mask=open("mask.png", "rb"),
    prompt="用中文翻译上面的文字，不改变其他内容。",
    n=1,
    size="1024x768"
)

# Save the returned image to disk
with open("output.png", "wb") as f:
    f.write(response['data'][0]['image'])

The prompt asks for Chinese and forbids other changes. n=1 requests one image. size="1024x768" sets the size in this example. The first returned image is written to output.png.

Review output.png. "Summer sale" should now read "夏季特卖," with a similar face and the same background. A second fixture for this class of model is Korean in and Vietnamese out: same picture, different glyphs. Compare files, not memories.
Save it as the deliverable only after that look. If position or weight is slightly off, nudge it in an editor instead of regenerating the whole frame.

Caveats. Proofread every string, including type you thought the mask excluded. A missing line means a tighter prompt or a smaller block. The API bills by image size and by generation count. If Image.create_edit is not the symbol in your installed SDK, keep the contract — orig.png, mask.png, that prompt, one image, an explicit size — and map it to the current client.

Tutorial 2: A hosted tool, exercised like a test

Overview. Use ReWords, or a similar site such as EditTextImage, to run Method 4 on one fixture. This walkthrough uses ReWords. No local install. The test is whether you can explain the output.

Prep. Open the site. Have one image whose source text you can read yourself. If you cannot read the source, you cannot grade the translation.

Flow. Upload. The system detects and OCRs. It shows a suggested translation. You edit and lock. It renders. You download and compare.

Upload. Choose the function that translates text in the picture and upload the fixture, an ad that contains "SUMMER SALE." Learn the clicks on a short banner before you try a menu or a comic panel. Short text makes a bad read obvious.
Pick the language and edit the strings. Detection should return every line. Select the target. Chinese is the example. Correct mistranslations in the suggestion list before you generate. Lock tokens that are not translations: brand names, model numbers, a price already in the target currency.
Generate. The render removes the original glyphs and draws the translation back in the original color, on the original background, in a face that tries to match. Same erase-and-redraw contract as the Python tutorial. You inspect the result as if you had watched the draw call.
Download and compare. In the example, one side still says "SUMMER SALE" and the ReWords result says "夏季特卖," with the package or banner background still in place. Flip them at 1:1. Look at edges, shadows, and any character that might have been invented.

How to judge the run. These products usually include a free allowance. ReWords includes several free generations per image. Past that you pay or you spend credits. That is enough to learn the tool and not enough to pretend a catalog is free. Quality on this class of banner is often high, and it is still a draft until someone checks numbers and proper nouns. The method earns its place when you cannot get the design source and you still need a localized image rather than a caption under the old picture.

If you need the same loop named as a product action, edit text in an image is the job the upload-and-generate flow is for. The acceptance strings are the spec: "SUMMER SALE" to "夏季特卖," and, on the EditTextImage-style banner, "Summer sale — 30% off everything" to "Sommer-Sale — 30 % auf alles." There is no screenshot in this article on purpose. Grade the strings and the seam, not a caption under a missing figure.

Tutorial 3: Python, Tesseract, and a redraw you can debug

Overview. A script reads text off an image, translates it, and paints the translation back. The stack is Tesseract, OpenCV or Pillow, and a translation client. The fixture is Spanish into English. The example output is translated_image.png.

Flow. Load the original. Detect and OCR with Tesseract. Call a translation engine — here, the Google Translate client in googletrans. Clear the original region with a draw. Paint the English with Pillow. Save. The first version uses one known rectangle so the draw call is visible. Production uses the boxes OCR returns.

Install the binary and the libraries. The Python package does not include Tesseract. On a Debian-family machine:

# Install Tesseract (example for Ubuntu)
sudo apt-get install tesseract-ocr

# Install Python libraries
pip install pytesseract pillow googletrans==4.0.0-rc1

The pin on googletrans==4.0.0-rc1 matches the Translator class used below. On macOS the Python packages stay the same and the system package does not. Do not skip language data. lang='spa' fails in a boring way when the Spanish traineddata is missing. If you later swap in an official translation client, keep the OCR, the rectangle erase, and the measured draw. Replace only the client.

Run the script. The image is image_with_text.png. Recognition is Spanish. The destination is English. The rectangle is a stand-in, not a detector.

from PIL import Image, ImageDraw, ImageFont
import pytesseract
from googletrans import Translator

# Load the original image
image = Image.open("image_with_text.png")

# Run OCR (recognize Spanish text)
text = pytesseract.image_to_string(image, lang='spa')
print("Recognized text:", text)

# Translate the text
translator = Translator()
result = translator.translate(text, src='es', dest='en')
print("Translation:", result.text)

# For this example the text position is known; in production you would
# use pytesseract.image_to_boxes to get a box per character.
x, y, w, h = 50, 50, 300, 50  # example coordinates

# Erase the original text on the image
draw = ImageDraw.Draw(image)
draw.rectangle(((x, y), (x + w, y + h)), fill="white")  # cover with white

# Draw the translated text
font = ImageFont.truetype("arial.ttf", size=36)
draw.text((x, y), result.text, font=font, fill="black")

# Save the output
image.save("translated_image.png")

What the script actually did. Tesseract reads Spanish from the whole image and prints it. googletrans translates es to en and prints that too. The rectangle x, y, w, h = 50, 50, 300, 50 is a known stand-in: origin (50, 50), width 300, height 50. Pillow covers it with white, draws the English in arial.ttf at 36 pixels, and writes translated_image.png. One fixed region, one flat fill, one face you did not match. That is the illustration, not a localizer.
What production changes. Take boxes from pytesseract.image_to_boxes, union them into a line, and erase the line rather than each character. Sample a nearby color instead of hard-coding white. Measure width with Pillow and wrap or shrink when the string no longer fits — German often will not, by 30% or more. Loop for multiple lines. Use OpenCV contours when the boxes are noisy. Layout fidelity still trails an inpainting model, so review one image before you batch. A useful split is JSON of boxes and strings, then a second process that renders, so a bad name is caught before it is painted across a catalog.

Pick From the Constraint You Actually Have

Choose with three facts: how faithful the file must be, what the work may cost, and whether anyone can still open the source. A preference for a vendor is not a fact.

No source file, and the calendar is short. This is the common case. Use an online AI tool in the Method 4 family, one you can use to translate the text inside an image without rebuilding the design by hand. Upload, choose the language, correct and lock lines, generate, download, then proofread. The render itself is seconds to a minute, plus the review you owe the file.
The image is text-heavy and structured. Charts, manuals, diagrams. Use OCR plus typesetting (Method 2). If the job is a pile of PDFs or scans, a format-preserving document service (Method 3) can own the reflow. A 60-page PDF in minutes is a Method 3 story. A labeled schematic with leader lines is a Method 2 story, because the arrows will not survive a blind reflow.
The file is print-bound or brand-critical. Use manual design (Method 1) or a translator plus DTP (Method 9). These are the slow, expensive, very-high-fidelity rows. Pick them when a visible patch costs more than the labor.
The work has to live inside a system you run. Use a multimodal API (Method 6) when a prompt-and-review loop is acceptable and the images may leave the network. Use a custom Python pipeline (Method 8) when you need batch control, a private network, or a box list you can store. Use diffusion inpainting (Method 7) when the background is the hard part and the text is short enough not to shatter.
You only need to understand the picture. Use a phone camera app (Method 5). A few seconds. Not a brand review. Overlay, default font, possible watermark or patch. Reading only. Never the file you publish.
The editable source is still available. Edit it directly (Method 10). This is the cleanest path that exists. The moment the file is a flattened export, you are on Method 2, even if someone named it like a source.

When two constraints conflict, fidelity wins on anything public and durable, and speed wins on anything internal and temporary. A slide for tomorrow can be a Method 6 draft. A package that will be printed cannot.

Five Habits That Keep the Layout From Slipping

Plan for length before you draw. German and French commonly expand by 30% or more. Chinese and Japanese commonly contract. Resize, re-wrap, or re-track. Never assume the new string fits the old box. Measure the rendered width and fail on overflow instead of clipping glyphs.
Match the type, not only the language. Reproduce weight, color, letter-spacing, and any stroke or shadow. A correct sentence in the wrong face still looks wrong. If you lack the identical font, record the substitution where a reviewer will see it.
Rebuild the background until the seam disappears. Feather the edge and check it at 1:1. On a script, leave the tutorial's white rectangle behind once the background is not white. Texture that suddenly goes smooth is an erase that did not heal.
Account for what is not a paragraph. Arrows, callouts, numbering, and legends carry meaning or point at it. A pipeline that returns only strings will skip them. Put them on the review list.
Proofread the words, not only the pixels. Machine translation is fast and imperfect. Numbers, units, brand names, and proper nouns need a human check before you publish. A clean fill around a wrong price is worse than a visible seam around a right one.

Keep the source image, the box list, the glossary, and the output in one job record. Otherwise a retired claim walks back in through a cached render.

What You Should Ship

Match the method to the file, not to the demo that was easiest to click.

Labels you can batch are Method 8: detect, OCR, translate, erase, redraw, seconds per image, and a fill that may be flat color or a neighborhood average. Diagrams with leader lines are Method 2, seconds to minutes for the read and the translation, then minutes to tens of minutes of typesetting. A long paid document is Method 3: fidelity described as 90% or better, a 60-page PDF in minutes, handwriting and strange layouts left out. A poster, package, menu, or comic with no source is the hosted repaint, free generations on a fixture and credits after that. "SUMMER SALE" to "夏季特卖" on the same background is a pass. An InDesign or Illustrator file is Method 10. A flatten is Method 2. A flagship, a trade-show wall, or a legal artifact is Method 1 or Method 9 — highest fidelity, highest time, highest cost.

A multimodal call is the coffee-machine diagram in Spanish, or an infographic whose words move and whose geometry stays: seconds to tens of seconds, and a miss when a line is skipped, misspelled, or far longer than the box. Diffusion is ten to tens of seconds, stronger on texture, weaker on long words. A phone app is a few seconds of reading and not a master.

Detection boxes are how you know which pixels may change. Pick the owner of that loop whose failure you can see before the file ships, and proofread the words before you trust the redraw.

Replace Raster Glyphs on an Exported Flyer: Inputs, Branches, and Failures

baiwei — Thu, 08 Oct 2026 12:37:42 +0000

A finished flyer is not a document you can edit. It is a bitmap, sometimes wrapped in a PDF container, and the words you want to change are paint. The date moved, the price changed, or a headline typo survived review. The PSD is gone. The Canva link is gone. The person who built the file may be gone too. Rebuilding the whole design is the expensive default, and it is usually the wrong branch.

Treat the job as a router. Name the input. Name the edit. Name the failure you will not accept. Then pick one of ten paths. A solid-color title block and a full-bleed photograph are different programs. Spending a Photoshop subscription on a five-second wording change is waste. Sending a dense texture through a one-click fill and hoping the scar disappears is also waste. The rest of this runbook is the branch logic, the procedure inside each branch, and the condition that should make you abort or switch.

The output you want is a new bitmap or PDF whose unchanged regions still look like the source, and whose changed strings say the new thing. You do not get the original layers back. Every method below is a different way to fake that recovery well enough for the medium you are shipping to. Almost nobody in this situation needs the original design file. They need the path that matches the flyer that is actually on disk.

Name the Artifact Before You Edit It

Start with the file you actually have, not the file you wish you had. Write four labels at the top of the job before any tool opens: file class, background class, edit intent, and the failure you refuse to ship. If you cannot fill those four labels, you are not ready to edit. You are still guessing.

Input classes

A JPEG, PNG, WebP, or TIFF. Text, logo, photo, drop shadow, and background are one grid of pixels. There is no text object to select. JPEG is the common handover, and it has already thrown away high-frequency edge detail, so the fringe around each letter is mushier than it looked in the design app.
A PDF that still contains real text objects. You can drag across a headline and the characters highlight. That is a different program from a scanned page.
A PDF that is only a placed image. It looks like a flyer in a PDF viewer and behaves like a JPEG. Do not trust the extension.
A vector export that was flattened on the way out. Illustrator or CorelDRAW may have been the authoring tool. You do not have the AI or CDR. You have the picture of it.

Edit intent

Write the change down before you open a tool. Replace string A with string B. Or remove string A and leave the background. Or change several strings that must stay aligned to the same grid. Length matters. A three-letter price and a two-line disclaimer do not stress the background the same way. If the new line is much longer or much shorter than the old one, say so now. That single fact knocks several "just retype it" paths out of the running, because the repaired hole will not match the new measure.

Also decide remove versus replace as a boolean, not as a vibe. Remove means the replacement string is empty and the only success test is the background. Replace means the new glyphs have to survive the same glance as the old ones. Mixing those two tests is how people approve a clean hole and then hate the new word, or approve the new word and ignore a scar around it.

Constraints you must record

Destination. Screen, social, or print. Print usually wants about 300 ppi so the new type is not softer or harder than the rest of the sheet. A social crop can tolerate a softer edge that a press PDF will punish.
Budget. Zero, a free browser tier, or a Photoshop subscription in the rough range of $20–$50 a month, with a trial available. Do not invent a cheaper plan to make the route sound easier. If the seat is not there, you are on a free branch.
Time. A clean background in the browser is usually under a minute. A complex photographic salvage is about 40 minutes of sampling and painting, not a two-minute typo fix. Put the number on the ticket so nobody schedules the hard branch into a five-minute gap.
Fidelity bar. "Close enough for a chat screenshot" and "this goes to a commercial printer" are different exit tests. Name which one you are being paid, or blamed, for.
Forbidden inputs. This work is not for IDs, invoices, official documents, or design source files. If the asset is one of those, stop. Do not route it through a cleanup tool to make it look editable.

The probe that classifies a PDF

Open the PDF. Try to select the headline with the text cursor, not the hand tool. If characters highlight and you can copy them, you are holding vector or live text, and the PDF path is the first branch to try. If the cursor never finds a text run, the page is an image. Converting that image "to PDF" does not create text. It creates a PDF-shaped bitmap, and you are back in the raster paths. Do this probe on every page you intend to touch. Mixed files exist: page one is live type, page two is a scan someone dropped in. Those are two jobs.

The probe that classifies the background

Zoom to 100 percent on the pixels directly behind the words you will change. Fit-to-window hides the exact texture that will betray the repair.

Flat color, or a color block with a hard edge: simple.
A long, smooth gradient with no detail: simple to moderate. Gradients fail when the fill averages them into a flat slab, so look for a vignette before you call the field simple.
Wood, brick, paper fiber, fabric, foliage, a crowd, or a face: complex.
Type that is also a texture (chalk, brush, neon, heavy outline, extruded shadow): complex, even if the wall behind it is flat. You are reconstructing an effect, not only a hole.

Write that label on the job. It decides whether you are allowed to take the short path. If two text blocks on the same flyer have different backgrounds, they are allowed to take different paths. The headline on a red bar can go through a fast fill while the date stamped on a photograph goes through sampling. Forcing one method onto both blocks is how a simple job inherits the scars of the hard block.

The Three Failures a Flattened Flyer Always Creates

Once the designer exported JPEG, PNG, or PDF-as-image, every layer collapsed. Headline, body, logo, background, and shadow became colored dots arranged to look like letters. The software no longer knows a glyph from a leaf. That single fact produces three failures. Every path in this runbook is a different repair for the same three. If a tool claims to skip all three and you cannot name which one it actually automated, you do not understand the branch yet.

1. The glyph boundary is not a selection.

You have to decide which pixels are the letter. Cut inside the edge and a halo of the old word remains. The next person who zooms will read both strings at once. Cut outside the edge and you delete background you cannot sample back from anywhere else. Thin strokes, diacritics, and the gap inside a counter (the hole in an O, the bowl of a B) are where this fails first. Anti-aliased edges make it worse: the "letter" is a ramp of mixed pixels, not a hard mask. On a JPEG, that ramp is also full of compression mosquitoes. A rectangular marquee around the word is fast and often wrong, because it includes background you did not need to destroy and still misses a descender or a shadow offset to one side.

The practical test: after you remove the word and before you type the new one, look at the hole. If you can still read the old word, the boundary failed. If a neighboring rule, logo edge, or face has a bite taken out of it, the boundary failed the other way. Fix the hole before you typeset. New letters are a terrible concealment layer. They only hide the middle of the scar.

2. The font is no longer data.

There is no font name, no point size, no tracking value, and no layer style. You identify the face by eye, hunt a close match, install it if you do not own it, then set size, weight, color, letter spacing, and effects such as a stroke or a drop shadow. Experienced designers still miss this by a small margin. Weight is the usual tell. A correct family in the wrong weight looks like a different flyer. Tight headline tracking is the second tell. If you leave default spacing on type that was originally crushed together, the replacement announces itself.

Color is the third tell, and it is cheaper to fix than the font. Sample the old ink. If the old word had a one-pixel stroke or a soft shadow, a naked fill will look like a sticker even when the family is right. You are matching a treatment, not a Unicode string.

3. The background under the words is already gone.

Erasing text does not reveal a hidden layer. Those pixels were never stored. On a solid color, filling the hole is trivial: you sample the swatch and the repair is invisible because there was nothing to invent. On a photograph, a gradient with a vignette, or a busy texture, you are inventing the missing image. That invention is how a "two-minute typo" becomes about a 40-minute salvage. If you cannot invent that region convincingly, no amount of careful typesetting on top will save the job.

Hold those three failures in mind while you read the paths. A method that only solves one of them will leave you doing the other two by hand. OCR solves the string and abandons the look. Clone tools solve the background and abandon the font. A PDF editor solves all three only when the glyphs were never flattened. Automatic image replacement tries to solve all three at once and must be judged by the residual, not by the demo thumbnail.

Short Path: Browser Replacement on a Simple Background

If the background behind the target words is a solid or a gentle gradient, do not open a desktop editor first. Modern tools built to replace text on a finished image collapse detection, a font-like match, and background reconstruction into one upload. For flyers, posters, menus, labels, and covers that only exist as exported pictures, that is the path that usually ends the job.

Use a product aimed at this task when you want to edit text on a finished flyer without reconstructing layers. ReWords AI is the one specified for this runbook. It does not drop a fresh text box on top of the old word and hope you hide the difference. It detects the text already in the picture, removes it, rebuilds the surrounding background, and generates replacement text that aims to follow the original font, color, and size.

Input. PNG, JPG, or WebP. A straightforward flyer, not an identity document and not a layered source file. If you only have a PDF-shaped bitmap, export a PNG or JPG of the page first. Do not expect the short path to edit PDF text objects. That is a different branch.

Procedure

Open the tool in the browser. Nothing to install. You do not need a local font folder for the first pass, because the point of this branch is that the model aims at the letters already in the picture.
Upload the flyer. Wait until the preview is the real file, not a tiny placeholder. A downscaled preview can hide the exact block you need.
Read the detected blocks. Click the block you intend to change. If the detector missed a line, draw a box around that line yourself. Missed blocks are normal on low contrast, on type that sits on a photograph, and on decorative lettering. Drawing the box is part of the procedure, not a failure of the run.
To remove the text, leave the replacement empty. To change it, type the new copy. Do not paste a paragraph into a box that used to hold one word unless you have accepted a layout change. Keep the new string in the same script the design used unless the brief really is a translation.
Sign in. Generate. New users typically get one free edit after signing in. No credit card is required for that first edit. It is a single starting edit, not an open meter. If you have more than one flyer, assume later runs are outside that first edit.
Review at full resolution, not at the fit-to-window preview. Check the boundary, the weight, and the background separately. Download only after that review.

Why this branch wins. The whole pass stays in the browser. On a clean background it is usually under a minute. You are not sampling textures, installing fonts, or rebuilding a frame in a layout app. For the majority of flyers, which really do sit on simple backgrounds, this is the correct first execution, not a demo you try after Photoshop fails. Using a desktop subscription for that five-second case is the wasteful mirror image of using a one-click tool on a texture so dense it will scar.

Failure conditions — switch or stop

The result depends on image sharpness, font complexity, background texture, and how much text you changed. Solid and gradient backgrounds come out cleanest. If the source itself is a tiny, heavily compressed JPEG, no branch will invent a crisp outline that was never captured. Re-export from a larger master if you still have one. If you do not, lower the fidelity bar before you blame the tool.
A dense photograph behind the words is handled as best as possible. It will not always disappear when you inspect at 100 percent zoom. If you can still see a smear, a repeated motif, or a soft patch shaped like the old word, this branch missed. Move to a sampling path (clone and heal in Photoshop, or the same idea in GIMP) or accept a redesign.
The generated letters are not pixel-accurate. Font style and exact edge rendering can drift. If the job is a logotype with a custom letterform, do not pretend the match is exact. Either accept the drift or redraw that one word by hand on top of a cleaned background.
Proof every character, including non-Latin copy. A wrong glyph is a failed run even if the background looks clean. Dates and prices fail this way: the texture is fine and the digit is wrong.
Abort entirely if the file is an ID, an invoice, an official document, or a design source file. The tool is not for those inputs. A design source file should be edited in the app that still has the layers, not flattened and then repaired.
If you need the original layers, styles, or a still-editable source for the next campaign, this path will not produce them. It produces a new finished image. Store that limit in the handoff so the next person does not ask you for the PSD you never recovered.

If the short path passes the 100 percent check, stop. Do not "improve" a clean result by running it through a second inpainter. A second pass is how you invent new artifacts. Ship the download, and keep the source file next to it so you can rerun from the original if the copy changes again. Rerunning from your own previous output stacks compression and repair error.

Path 1: Content-Aware Fill, Then a New Type Layer

Tools. Adobe Photoshop, paid subscription, roughly $20–$50 a month, trial available.

Input. Any bitmap: JPG, PNG, or TIFF. Best when the words sit on a flat color or a smooth gradient, such as a solid title block, and you are replacing a small amount of type.

Preconditions. You can already make a selection and use the Type tool. You do not need to be a retoucher. You do need the subscription. If the trial is what you have, do the job inside the trial window and export before it ends. A half-finished PSD you cannot reopen is not a deliverable.

Procedure

Open the flyer. If the background is locked, double-click the layer so you can stack a real type layer above it. Do that before you forget, or the new text gets baked into the only pixels you have and you cannot nudge it.
Select the text area with the Rectangular Marquee or the Lasso. Include the anti-aliased fringe and any shadow that belongs to the old word. Exclude neighboring letters you intend to keep. For a single bad word inside a sentence, the Lasso is safer than a rectangle that eats the words beside it.
Choose Edit, then Fill. Set the fill type to Content-Aware. Let Photoshop rebuild the background behind the text.
Deselect. Inspect the patch at 100 percent before you type. If the fill already looks wrong, do not hide it under new letters. Undo and reselect. Feather the selection edge with Shift+F6 when the blend line is obvious, then fill again.
Switch to the Type tool and set the new string in the same spot.
Sample the original text color with the Eyedropper. Match a similar font and size. Copy stroke or shadow treatments if the old word had them. Layer styles are not recovered automatically from a flattened file. You recreate them.
Nudge position and layer order. The new type should sit in the same stacking place as the old word, not above a logo that used to cover it, and not underneath a photo that used to sit behind it.
If Content-Aware left a repeated speck or a blurry island, clean it with the Healing Brush at low opacity. Low opacity is the point. A full-strength dab paints a new scar.
Export at the resolution you actually need. For print, stay near the source, usually about 300 ppi. Upscaling a small JPEG at export time does not create the detail the fill failed to rebuild.

What this path solves. Boundary, by your selection. Background, by Content-Aware Fill, which is strong on flat color and clear gradients and produces a natural transition there. Font, entirely by you.

Failure conditions

Busy texture: the fill repeats a motif or goes soft. Small type and fine detail leave traces of the old strokes. That is the cue to leave this path, not to run Fill five more times on the same pixels.
A selection that is too tight leaves a rim in the shape of the old letters. Expand it slightly and include the fringe. A selection that is too wide eats a border, a face, or a second line of type.
Content-Aware Fill does not know your new sentence. If the new string is longer than the hole, it will cover unrepaired flyer. Either repair a larger region first or shorten the copy.
This path does not remove the need to match weight, tracking, and color. A perfect fill under the wrong type still fails the exit check.

Difficulty. Medium, if you already know basic Photoshop. In the comparison matrix that is two stars, not four. Cost. The subscription above, on the high side versus free tools. Time. Medium, not the one-minute browser pass. Formats. JPG and PNG in the matrix. TIFF is also a normal Photoshop bitmap for this fill, so do not refuse a TIFF just because the summary column says JPG/PNG. Fidelity. Three stars when the background matches what this path is for. It is not the most faithful repair in the toolkit, and it should not be sold as one.

When to refuse this path. Portraits, wood grain, brick, and other photographic backgrounds. Those belong to sampling. Also refuse it when you do not have Photoshop and will not subscribe.

Path 2: Clone, Heal, or Patch, Then Retype

Tools. Photoshop again. Clone Stamp, Healing Brush, or the Patch tool. Same subscription as Path 1.

Input. A bitmap whose background is a real texture or a photograph: wood, brick, a portrait, a complex gradient, handwritten or artistic type. This is the path for backgrounds that Content-Aware Fill smears.

Preconditions. You are comfortable sampling. Beginners leave repeating stamps that look like a rubber stamp of the neighboring pixels. If that description is you, do not start a client delivery on this path today. Practice on a copy, or hand the texture to someone who already has the hand, or accept a lower fidelity bar on another branch.

Procedure

Open the flyer. Duplicate the background layer before you paint. The duplicate is your rollback. Flattening your only copy and then painting is how you lose the job.
Select the Clone Stamp or the Patch tool. Hold Alt and sample clean background beside the text, not from a region that already contains a different edge, a highlight from the wrong side, or a logo.
Paint over the words. Follow the direction of the texture or the color block. On a brick wall, follow the courses. On wood, follow the grain. On a face, stop and reconsider. Skin is where this path embarrasses people.
Keep the brush large enough to cover a stroke and soft enough that the edge does not draw a circle. Lower the hardness. Work in passes, not one opaque smear. Change the sample point often so you do not tile one lucky patch across the hole.
The Healing Brush is the better default when color has to blend as you cover the text. The Clone Stamp is the better default when you must preserve a hard structural line the healer would melt. The Patch tool is useful when the region is a coherent block you can drag onto a clean neighbor. Use the one that matches the texture, not the one you opened first.
Deselect or release the tool once the old word is gone. Check for tiny repeated blocks and color shifts at 100 percent.
Make a new type layer. Match font and color with the Eyedropper, then use a color adjustment if the new word still floats. Set the new copy. Tune position and opacity so the type sits in the scene instead of hovering over it.
Export from the repaired file. Keep the duplicate layer until the client has accepted the pixels.

What this path solves. Background reconstruction on complex patterns. The Healing Brush blends color as you cover the text, so the patch can look natural if your samples were honest. The font is still your job.

Failure conditions

Large areas need many passes. You will see repeated tiles if you sample one small spot and spray it everywhere. When you see a tile, stop and resample. Do not cover the tile with type and call it done.
Faces and skin tones shift easily. If the old word crossed a cheek, an eye, or hair, budget the full salvage. This is the branch where a two-minute typo becomes about 40 minutes.
You cannot rush the sample. A sample taken from the wrong side of a highlight will paint the highlight into the shadow.
Seamless work is precision work. If you need the file in five minutes and the background is a crowd photo, this path will not meet the clock. Switch to the short path only if you can accept a visible repair, or to a full regeneration if you can accept a redesign.
Working on the only layer, with no duplicate, means a bad stroke can become the new source. That is an operator failure, not a tool failure.

Difficulty. High. The matrix gives it four stars. Time. High. Cost. Photoshop subscription, on the high side relative to free tools. Formats. Any bitmap. The matrix lists JPG/PNG. Fidelity. Four stars, higher than Content-Aware Fill, which is why the slower procedure exists. Do not invert that. Free is not "better." Slow is not "worse" when the texture is the product.

Path 3: Generative Fill on Photoshop 27.10 or Newer

Tools. Photoshop 27.10 or newer. That version is the floor for Generative Fill in this runbook. Adobe Firefly is the model family included with the Photoshop subscription. Third-party plugins such as StartAI or Luminar AI can sit beside it. Plugins may have a free tier and a paid tier. StartAI's basic features are free. You still need a network connection, and the generative step spends credits.

Input. A bitmap, or a file that still has usable layers. The ask is a wording change you can describe in a short prompt, including larger type or a complicated scene where painting by hand is the wrong spend. This is a speed branch with a high fidelity ceiling relative to a casual fill, not a promise that every pixel will match.

Procedure

Open the flyer. Confirm the version is 27.10 or newer before you hunt for the command. On an older seat the menu item is simply absent. That is a version failure, not a selection failure.
Select the whole image or only the target region. A tighter selection gives the model less room to redraw a logo, a face, or a QR code you needed to keep. A half-selection plus a precise prompt is often better than selecting the entire poster and asking for a small wording change.
Right-click or choose Generative Fill, or Prompt to Edit, on Photoshop 27.10 or newer.
Describe the change in words. The useful shape is short: replace the text A with B. If the source string is Chinese, say that you are replacing that Chinese text rather than hoping the model infers the script. Keep adjectives out unless they constrain color or position.
Pick a model, such as Firefly. If you also have StartAI available, generate on both and compare. One of them will usually keep the surrounding layout better. You cannot know which before you look.
Generate several variations. Do not accept the first frame because it loaded fastest. Reject any frame that moves elements you did not name.
Choose the best composite. Photoshop puts the repaired background and the new text together. If color or shadow still drifts, fix those yourself. The model will not reliably copy a layer style it was never given.
Proof the actual characters at 100 percent, then export.

What this path solves. All three failures in one automatic pass, with more than one candidate, which is why the time rating is low. It is usable by someone who is not a retoucher. It can take on large type and complex backgrounds that Content-Aware Fill smears. Difficulty stays in the low-to-medium band: two stars in the matrix, because basic Photoshop plus an automatic step is enough.

Failure conditions

Output is not pixel-accurate. Treat that as a hard limit, not a wording quibble. Font, layout, and small geometry can drift away from the source. You do not get to claim the pixels match. "High realism" in a thumbnail can still fail a logo at 100 percent.
Chinese-language support varies by model. Some runs need an English prompt or a manual correction after the fact. Artistic lettering is not guaranteed. If the characters are wrong, the variation is trash no matter how pretty the paper texture looks.
A big change can take several prompt attempts. If the third variation still redraws the logo or the QR code, stop prompting and mask more tightly, or leave this path.
You need the 27.10 or newer baseline and an Adobe subscription for the native command. A plugin does not erase that floor when the command you are using is Photoshop's own Generative Fill. Firefly rides with the subscription. A plugin's paid tier is a separate medium cost, which is why the matrix allows "subscription (high) or plugin (medium)."
Results are unpredictable. Friendly to non-experts does not mean safe for a print deadline you cannot miss. Generate, compare, then decide.
No network, or no remaining credits, means this path is closed. Fall through to clone and heal or to a browser tool. Do not keep clicking a grayed-out generate button.

Difficulty. Low to medium. Cost. Adobe CC subscription on the high side, or a plugin at a medium cost. Formats. Bitmaps, or files with usable layers. The matrix says JPG/PNG. Time. Low, because the fill itself is automatic. Your time goes into rejecting bad variations, not into painting. Fidelity. Four stars is the ceiling rating, not a guarantee for the variation you happened to click. Pick the frame that earns it.

Best for. Fast passes over e-commerce and creative posters, complex scenes, and operators who want speed more than a hand-built match.

Path 4: GIMP, When the Budget Is Zero

Tools. GIMP, free and open source, from gimp.org. No subscription. This is the zero-budget desktop branch.

Input. A common bitmap. The job is a small edit on a simple poster, or you specifically need Windows, Mac, and Linux without buying Photoshop. If the file is a complex photograph and you do not have painting skill, free will not save you. It only removes the license blocker.

Procedure

Open the flyer. Duplicate the layer you will paint on. GIMP's history is not a strategy.
Choose the Clone or Heal tool.
Alt-click a clean patch of background to sample. Sample close to the hole so the lighting matches, but not from inside the old strokes.
Paint over the text region until the old word is gone. Follow the texture the same way you would in Photoshop. The brush physics are less forgiving, so check more often.
Use the Text tool. Create a new text layer. Match font and position by eye. Set opacity and alignment on that text layer rather than merging early.
If the edge of the paint looks harsh, smooth it with the Blur tool. A light blur on the repair fringe only. Do not blur the whole poster.
Optional: install the Resynthesizer plugin when you want behavior closer to a content-aware fill. It is still not Photoshop's fill. Do not expect the same success rate on a texture, and do not skip the visual check because a plugin ran.
Use Layer Copy (Ctrl+Shift+D) often enough that a bad stroke is undoable as a layer, not only as a history step you might have overwritten.

What this path solves. The same manual reconstruction as cloning in Photoshop, without the subscription. Layers exist. Import and export cover a wide set of formats. PSD compatibility is limited, so do not promise a round-trip of Photoshop layer styles. Some layer effects are simply not there. If the client hands you a PSD full of effects, open it somewhere that understands those effects, or accept that GIMP will flatten the look.

Failure conditions

Automatic repair is less capable than Photoshop's. Quality is your painting, full stop. There is no hidden smarter pass waiting in a menu you have not found.
Matching fonts and effects is manual. There is no style token stored in the flattened file for GIMP to read.
Time stays high. Free does not mean fast. The matrix marks time as high, difficulty at three stars, fidelity at three stars, cost free. Keep that shape. Do not promote this row to the top of the fidelity column because the price is zero.
If you needed a one-click result and you opened GIMP, you picked the wrong branch. Go back to the browser replacement or to a one-click inpainter.
Cross-platform is real: Windows, Mac, and Linux. That is a reason to choose this path in a mixed shop. It is not a reason to choose it for a face retouch you cannot already do.

Difficulty. Medium to high. You must know GIMP's basics. Formats. Common bitmaps, plus many import and export options. The matrix summarizes them as JPG/PNG. Best for. Zero-budget edits and cross-platform work. Not for a commercial printer who will inspect a face near 300 ppi unless you already have the painting skill the Photoshop clone path assumes.

Path 5: Photopea in the Browser

Tools. Photopea, a free browser editor that can open PSD-style files along with JPG, PNG, AI, PDF, and more. Most ordinary features are free. AI features need Photopea Premium or a third-party API such as Dezgo. The free tier is limited. Do not discover the paywall after you have already talked the client through a Magic Replace workflow you cannot run.

Input. You cannot install Photoshop, or you will not. The file might even be a PSD you were handed without a Photoshop seat. The edit is an occasional small text swap, or you need PSD-like layer access on a machine you do not administer.

Procedure

Open the editor in the browser and import the flyer. Wait for the full file. A large print board can lag. If the tab is struggling, export a cropped region of just the text and repair that, then composite back, rather than forcing the entire 300 ppi sheet through a tired browser.
Work in the Photoshop-shaped way. Use a selection tool, then Edit, then Content-Aware Fill, to remove text. That fill is still in beta. Plan on manual touch-up. It is not as strong as Photoshop's Content-Aware Fill. Complex areas need a manual patch. Budget the patch. Do not budget a miracle.
Or use the Healing tool and paint the words out yourself, with the same sample discipline as the desktop clone path.
When the old text is gone, add the new string with the Type tool (T). Match size, color, and tracking by eye against the original, which you should keep open beside the edit.
If you use Magic Replace, sign in. Usage limits apply. Enter an English prompt to swap the text. Review the frame before you trust it. If you are logged out, Magic Replace is not a step you can execute. Fall back to heal plus the Type tool.
If the repair looks speckled, try Filter, then Despeckle, lightly. One pass. A second pass starts to melt edges you wanted.
For Chinese type, pick a font the system actually has, then nudge letter spacing until the line matches the visual density of the original. Average recognition on the AI step means you proof the string yourself. Do not assume the prompt's Chinese came back intact.

What this path solves. No install, broad format support, layers and masks, and an interface people who know Photoshop can learn quickly. It is the right branch when the blocker is the machine, not the complexity of the texture. Difficulty is low to medium, two stars. Time is medium. Fidelity is three stars. Cost is free for the basics. That is a real advantage, and it is not the same advantage as five-star PDF text.

Failure conditions

Content-aware fill here is weaker than Photoshop's. If the beta fill leaves a blotch on a photograph, switch to the Healing tool inside the same session or move the file to a desktop clone path. Do not repeat a beta fill and hope the average improves.
The browser session depends on the network. Large files lag. A high-resolution print board can be miserable in a tab. If the brush trails seconds behind your hand, you will make bad samples. Stop.
AI results depend on the model behind Magic Replace. Chinese recognition accuracy is average. Proof it character by character.
A paid upgrade or an API key is a real gate for the AI features. The basic editor being free does not make the AI step free.
You can open many formats, including PSD, AI, Sketch, JPG, PNG, and PDF. Opening is not the same as preserving every Photoshop effect. Check the layers you care about before you promise the client the file "came across fine."

Pro move from the field. Roughly clear the text in some other pass first, then do the last edges here, if the file is large enough to fight the browser. The browser editor is a finisher for small swaps more often than it is the place to reconstruct a billboard.

Path 6: Edit the PDF, and Only If the Glyphs Are Still Text

Tools. Adobe Acrobat Pro, Foxit PDF Editor, WPS PDF or WPS Word, and Inkscape. Inkscape is free and can edit PDF, SVG, and AI. Acrobat Pro is expensive. Foxit's advanced tier is paid. WPS PDF's free tier is usually sufficient for a straightforward text change. Pick the app you already have that can actually enter text-edit mode. Do not buy Acrobat for a one-line price change if WPS can already select the line.

Input. A flyer that is already a PDF. The high-fidelity outcome exists only when that PDF still holds real vector text. Five-star fidelity in the comparison table means that case, and only that case. A scanned page or a PDF made by dropping a JPEG onto a blank page does not inherit those five stars. If you skip the probe and edit anyway, you will either be refused by the tool or you will be typing into an invisible layer while the old raster word still shows through.

Procedure

Run the text-cursor probe. If the headline highlights, continue. If it does not, leave this path. Do not convert an image into a PDF in order to qualify. That produces a container change, not text.
Open the PDF in the editor. Use Edit Text. Change the string, and if you must, the font or style. Change as little styling as you can. Every unnecessary font change is a new way to miss the original.
If a font was not embedded, the editor will substitute. A substitution is a visible failure even when the words are correct. Install the face or pick the closest face that is actually available, on purpose. Do not accept a silent swap to a default serif.
Exit edit mode so the styling locks again. Save, or save a copy, so the incoming file remains the rollback. Prefer a high-quality PDF export so you do not add a needless compression pass. Compression is how a clean vector headline becomes soft on the next hop.
Alternative inside this family: export the PDF to Word or InDesign, edit, and write a PDF back. That is useful when the change is structural. It is also how layout drifts. Or open vector text in Inkscape, which is often the easiest free way to reach real text objects and to edit SVG or AI siblings of the same artwork.
WPS can convert PDF to Word when you want the strings in a writing app quickly. Foxit's OCR can turn a scan into editable text, but the moment you rely on OCR you have left the five-star branch. You are now doing recognition plus reflow, and you should judge the result by the OCR path's limits, not by this path's stars.
For a multi-page handout, repeat the probe per page, edit the live pages in place, and route the scanned pages elsewhere. Saving the whole file out of a Word conversion because one page was hard is how you wreck the pages that were already perfect.

What this path solves. When the text is actually text, you keep layout and sharpness because you never rasterized the glyphs. You can also annotate or highlight, but annotation is not the edit. The edit is the string change. Difficulty is two stars. Time is low to medium. Cost is partly free, for example on WPS. Formats are PDF and, if you exported, Word. That combination is why this row can sit at five stars while cheaper raster tools do not.

Failure conditions

Image-only PDFs need OCR before any honest edit. Pretending otherwise wastes the afternoon.
A large change reflows the line and collides with a photo, a rule, or a page number. The layout can break even though the text engine is real. If the new string is much longer, reduce the point size deliberately or rewrite the copy. Do not let the line crash into the logo and then "fix" it by converting the page to an image.
Missing fonts force a substitute. That substitute fails the "same flyer" test. Embed fonts on the way out if the printer needs to open the file somewhere else.
Bulk edits across a multi-page document are a reason to stay here, but only after the probe succeeds on each page you will touch.

Fidelity. Five stars only for real vector text. Say that every time you quote the table. A scan edited after OCR is not this row.

Path 7: OCR the Words, Then Typeset a New Layout

Tools. An OCR engine or service: ABBYY FineReader, Tesseract, Baidu OCR, Tencent OCR, or the image-to-text path in WPS or OneDrive. Then a layout app: WPS Word or PowerPoint, Adobe Illustrator, Canva, Inkscape, or whatever you already design in. OCR often has a free tier, including mobile and online options, and open-source Tesseract is the usual zero-license engine. The design app is a separate cost. Office, WPS, or Inkscape can keep that second half free.

Input. A text-heavy flyer with a regular layout: a promo sheet, a simple brochure. You can accept some design drift. You need the wording changed more than you need the original shadow style. If the value of the flyer is the photograph, this is the wrong branch. You would be rebuilding a poster in order to change a caption.

Procedure

If the scan is noisy, denoise it in Photoshop or GIMP before OCR. Garbage in the bitmap becomes garbage in the text run. A high-quality scan or a clean export beats a clever engine on a muddy photo.
Run OCR. Read every line against the flyer before you design anything. Recognition errors are normal. Chinese recognition is a strength of several of these engines, and you still proof it. A wrong digit in a price is a failed job that looks finished. Dates, phone numbers, and names get a second human pass. Do not skip them because the paragraph "looked right."
Create a new document at the same pixel dimensions or the same page size as the source. Build a matching background. You are not editing the old file. You are reconstructing a cousin of it. Start from the original's measurements so the cousin at least has the same frame.
Paste the corrected text. Set font, size, color, and alignment. Use a face close to the original. If you have Photoshop's Match Font available, use it as a hint, not as a verdict. Then tune tracking and leading. Default leading is the usual reason a reconstructed text block looks looser than the flyer.
Rebuild rules, photos, and logos as separate elements. OCR did not capture them as editable art. Leave a placeholder if you do not have the logo source, and do not trace a bad JPEG logo inside this branch unless you have also accepted the vector path's cost.
Save an image or a PDF. If the destination is print, set the new type at the resolution of that destination, about 300 ppi, so the only soft thing in the file is whatever photograph you placed, not the words you just retyped.

What this path solves. You get real, editable strings quickly, in an app you already know how to type in, and the new text can be sharp because it is fresh type, not a healed bitmap. That is valuable when the flyer is mostly words and a slight redesign is acceptable.

Failure conditions

The layout is rebuilt from scratch. Complex designs make this slow, which is why the time rating is medium and tagged as re-layout, not as a one-click. If the flyer has overlapping art, angled type, or a dense grid, the "medium" estimate is optimistic. Say so before you start.
You should not expect a pixel-accurate copy of the original. Shadows, textures inside the letterforms, and exact spacing are the first things you will fail to reproduce. State that limit up front. The rebuild is a new layout that says the right words, not a restoration.
Fidelity in the table is two stars. Difficulty is two stars. The work is conceptually straightforward and visually loose. Do not upgrade this row because you spent all day on it. Effort moved the time. It did not move the star that says the result diverges.
Photos and illustrations are a second project. Leaving a hole where the product shot was is not done. Either place the original crop or commission the art. Do not stretch a fragment OCR happened to ignore.

Cost. OCR is often free or open source. Word or an equivalent is enough for regular layouts. Formats. JPG, PNG, or PDF in. The matrix writes the output side as DOCX. From there you can go to PowerPoint, SVG, or AI if the layout needs it. Best for. Wording changes on regular, text-heavy layouts. Not for a poster whose value is the photograph.

Path 8: Inpaint the Background, Then Add Type Elsewhere

Tools. One-click cleanup and inpainting sites such as Cleanup.pictures, ClipDrop's web Remove, Pixelcut, Inpaint, and Img.LL. They are a family, not identical products. Try more than one on a difficult hole and keep the best plate. Do not average them by stacking.

Input. JPG or PNG is the format class you should count on. The task is to remove words, a stray object, or a watermark, then place new type. Some of these tools let you add text in the same session. Many do not, and you should assume you will set the new type in Photoshop or another editor afterward. The cleanup and the retype are two steps. Finishing only the first one is a missed delivery.

Procedure

Open the site and upload the flyer. Confirm you are looking at your file and not a sample asset the homepage left in the canvas.
Use the erase brush. Cover the whole text region, including the anti-aliased fringe and the shadow. A brush that only hits the stems leaves ghosts. A brush that slops over a face deletes evidence you needed for the reconstruction. Err a little past the fringe, not a lot past the subject.
Let the tool reconstruct the background. Download the full file it actually offers, not the preview screenshot.
Inspect at 100 percent before you typeset. Look for a soft rectangle, a plastic texture, or an edge that does not continue a line from outside the mask.
Add the new string in that tool or in a second app. Note the surrounding colors and sample them with an Eyedropper so the new type matches the repaired scene. If you guess the color from memory, you will miss it.
If a residue remains, finish with Photoshop's Patch or Clone tools. Do not stack three different inpainters on the same hole. Each one invents a slightly different texture, and the overlap looks worse than the first miss.
If you are comparing tools, use the same source crop and the same mask so the comparison means something. A flattering crop is not a test.

What this path solves. Speed. No Photoshop skill, a brush and a click, sometimes a Chinese interface, sometimes batch processing on a paid plan. Time in the table is very low. Difficulty is two stars. Fidelity is three stars: useful, not the top of the scale. Cost is free to start, with a paid fast tier for some features.

Failure conditions

Complex patterns will not be perfect. The repair can look slightly synthetic, with blurry edges on a large removal. That is a weaker result than a careful manual clone, and the matrix already says so. Switch rather than arguing with the preview.
Output resolution may be capped. Ultra-HD or an HD download often sits on a paid plan. If the print spec is about 300 ppi and the download is a small preview, the path failed the output contract even if the words are gone. Check pixel dimensions before you celebrate.
Some free tiers stamp their own watermark in the corner. You removed one mark and inherited another. That is not a finished flyer. Either pay for the clean download or choose a different tool. Do not ship someone else's bug as your design.
Stripping a watermark you do not own, for commercial use, carries copyright risk. The fact that a brush can cover it is not permission. If the mark is a stock-photo or template watermark, stop and get a legitimate file.
You usually still have to set new type somewhere else. An empty clean background is only half the ticket. Leave time for the type match. The "very low" time rating is about the removal, not about becoming a typographer.

Formats. JPG and PNG. If you were handed a PDF, rasterize a page only after you have confirmed it is not live text. Rasterizing live text in order to inpaint it is a downgrade you chose.

Path 9: Regenerate a Lookalike From a Prompt

Tools. DALL·E 3, which means a ChatGPT subscription, because DALL·E needs ChatGPT Plus. Midjourney, on its own subscription. Stable Diffusion through a WebUI and hosted variants. Regional tools such as Baidu's Wenxin Yige, some of which are free. Cost for this class sits in the medium-to-high subscription band, not at zero, unless you are on one of those regional free tools. Do not quote a price the source does not give you. "Subscription" is the fact. A made-up monthly number is not.

Input. You are willing to leave the original layout. You want a fast creative variant, or you are not attached to the exact styling. You can supply the original flyer to a service that actually edits images, such as DALL·E's edit mode. If the brief says "change only the date and touch nothing else," this path is probably dishonest. Read the brief again before you prompt.

Procedure

Upload the source flyer where the service supports image input. A text-only prompt that describes the flyer from memory will drift further. Use the picture.
Prompt in a tight shape: based on this poster, replace the original text A with B. Constrain the exact wording and where it sits. Name colors and the overall style. If you need print-oriented detail, you can say high-resolution print quality in the prompt. That phrase is a hint to the model, not a guarantee of a file at about 300 ppi. Check the exported pixel size yourself.
If the model supports a mask, erase the old text first and let it regenerate only that region. An unmasked run will redraw the whole poster, including pieces you wanted frozen.
Generate more than once. Expect several prompt-and-output rounds to get close. Save the prompt text beside the images so you can tell which constraint actually helped.
Composite QR codes, logos, and any legal line back by hand. The model will not faithfully keep them. A regenerated QR code that "looks like a QR code" is a broken code. Replace it with the real asset.
Proof the words, especially if the copy is Chinese. Comprehension can be off. A beautiful poster with a wrong character is a failed run.

What this path solves. A new string and a new overall style in one generation, without drawing. It is a redesign branch. Difficulty is low, two stars, if you can already use the service. Time is low, and it is generation time, not painting time. Fidelity is two stars, the same low band as OCR, for a different reason: the picture diverges on purpose. That low fidelity is not an insult. It is the contract. You asked for a new picture in the manner of the old one.

Failure conditions

Layout and font distort. If the output moved the photo, changed the logo, or reflowed the grid, and the brief was a wording fix, this path failed the brief. Do not ship it because it looks polished.
Quality is capped by the model. More adjectives do not remove that cap. Switching from one service to another is allowed. Promising that the third service will match pixels is not.
You must plan a manual composite for anything that has to stay exact. If you do not have the logo file, the QR asset, or the legal line as real text, regeneration has nothing correct to paste back.
Unpredictable output means you schedule rounds, not a single click, even though each round is fast.

Formats. The matrix lists JPG/PNG. In practice you are feeding the service an image plus a prompt and taking an image back. When this is the honest choice. You already decided a redesign is acceptable. It is not a fallback you pretend is a restoration.

Path 10: Trace to Vectors and Rebuild

Tools. Adobe Illustrator, CorelDRAW, or Inkscape. Illustrator and CorelDRAW are paid, subscription or license. Inkscape is free. Use Inkscape when the budget is the blocker and you can live with its trace controls. Use Illustrator or CorelDRAW when a printer or a brand team expects those files and you already have the seat.

Input. A bitmap that came from a vector design, especially if the logo or the icons must change too, and you are comfortable in a vector editor. High source resolution helps. A soft JPEG traces into mud. If the flyer is a photograph with a caption, do not trace the photograph. Trace is for shapes that want to become paths.

Procedure

Place the flyer image in the vector app. Do not assume Place creates editable type. It creates a linked or embedded picture. The type tool will not land inside those pixels.
Run Image Trace, or the equivalent command in your app, to convert the bitmap into paths. Prefer a crop of the logo or the text block over tracing the entire photographic page.
Expand and break the result into pieces you can select. Immediately inspect anchor points. A clean contour is a keeper. A cloud of nodes around a soft edge is noise.
Try to select the paths that used to be letters, and delete those if you intend to retype. Or ignore the traced letters and cover that area with a clean shape you draw yourself, then put live text on top.
Retype with the Type tool. You can reuse traced shapes that are actually the logo, if they traced cleanly. If the font was not traced into outlines and somehow remains live text, edit it directly. A bitmap trace rarely gives you that. Do not plan on it.
If the auto-trace welds the letters to a shadow or a gradient, stop tracing the whole page. Trace only a text crop. Use a high-threshold black-and-white preset when the art is really line art. Release the anchor points that are noise and keep the ones that describe the contour.
If you still cannot get a clean cut, use the Pen tool and redraw the contour yourself. Then smooth jagged paths. Jagged paths are the normal failure of a sloppy trace, not a style.
Adjust the layout. Save PDF or PNG for delivery. Save SVG or AI if the next person needs the paths. A PNG-only handoff throws away the reason you traced.

What this path solves. After a successful trace, graphics are paths you can keep editing, and new words are real text on top. Fidelity can be high, which is why the table gives this path four stars, not five. Five remains reserved for a PDF whose text was already real text. Time and difficulty are both high, four stars on difficulty. This is professional-user work. It is the right branch when the logo or the icons must change and the flyer really did start as vectors.

Failure conditions

Image trace is bad at separating type from artwork. Shadows and gradients fuse into the background. When you see the letter and its shadow as one black blob, the trace has already failed. Switch to the Pen, or abandon the path.
You get jagged edges and a pile of useless nodes. That is not a finished logo. Cleaning every node by hand can cost more than redrawing the mark. Choose redrawing earlier than your pride wants.
Multi-color gradients trace poorly. A flyer that is mostly a photograph should not be traced at all. Use clone and heal.
The procedure is slow. Starting here because "vectors are cleaner" is how an afternoon disappears. Cleaner is only true after a good trace. Before that, you have a worse file than the JPEG you started with.
Conversion can demand a high-quality source. If you do not have one, this branch's fidelity rating is theoretical. Do not quote four stars to the client while you trace a thumbnail.

Formats. Bitmap in. SVG, AI, or PDF out. The matrix names JPG/PNG on the way in. Cost. Paid for Illustrator or Corel, free for Inkscape.

Failure Signals and the Branch They Imply

After any path, read the picture before you read your own hope. These signals are observable. Each one maps to a next action. None of them requires a new product.

You can still read the old word through the new one. The boundary was too tight, or a shadow was left outside the mask. Expand the repair to include the fringe and the shadow. Do not add a second copy of the new word on top.
A hard rectangle of slightly wrong color sits behind the new line. Content-Aware Fill or an inpainter was aimed at a texture it cannot invent. Switch to clone, heal, or patch, and sample along the grain.
The same pebble, brick, or skin pore repeats like a stamp. The clone source was too small or never moved. Resample from several neighboring points. If the region is a face and the repeat is already obvious, budget the long salvage, about 40 minutes, or stop and tell the requester the photograph cannot be invisibly rewritten on this clock.
The whole poster moved: the photo is different, the QR is nonsense, the logo grew a new shape. You are in the regeneration branch and it did what regeneration does. Composite the real logo and the real code back, or leave the branch. Do not prompt again with more adjectives and expect the code to become valid.
The PDF text cursor never highlights. You are not in the five-star path. Stop editing the container and classify the page as a bitmap.
After a PDF text edit, the letters are suddenly a different family. The font was not embedded and the editor substituted. Install the face or choose the closest available face on purpose, then re-export with fonts embedded if the printer requires it.
A watermark appears in a corner you did not paint. A free tier of an inpainter signed the file. Pay for the clean export or switch tools. Do not crop the signature off and pretend the license allowed the edit.
A traced logo is jagged and has far too many anchors. Image trace lost the contour. Use the Pen tool on the important outlines, or stop. Do not send the noisy paths to a vinyl cutter or a sign shop.
Generative Fill or a text-to-image model mangled Chinese, or ignored the script. Language support varies by model. Retry with a short explicit prompt, or set the string yourself on a cleaned background. Wrong characters are a hard fail.
The new line is longer than the repaired hole and now covers a photo edge or a border. The edit intent was underspecified. Repair a wider region before you typeset, or cut the copy. The tool will not negotiate with the grid for you.
The file is an ID, an invoice, an official document, or a design source. There is no next branch. Stop, including when a previous click already produced a preview. Do not download and "just see."

Write the signal you actually see into the ticket. "Looks weird" is not an input the next path can use. "Repeated brick tile to the right of the date" is.

Comparison Matrix

Read the stars in one direction only. More filled stars means harder, in the difficulty column, or closer to the original design, in the fidelity column. Difficulty tops out at four filled stars. Fidelity reaches five filled stars on the PDF row alone, and only when that PDF still contains real vector text. An empty outline star is not a filled one. Do not reorder the rows because a tool is newer or cheaper. Fidelity means how closely the result can resemble the original design, not how new the feature is. Time follows complexity. Cost is the subscription or the free option, not a moral score. A free row is not the most faithful row. The PDF row is the most faithful row only when the text is still real vector text. AI rows do not become more accurate because the time cell says low.

Method	Difficulty	Fidelity to original	Time	Cost	Input formats
PS Content-Aware Fill	★★☆☆☆	★★★☆☆	Medium	Paid subscription (high)	JPG/PNG
PS Clone/Heal	★★★★☆	★★★★☆	High	Paid subscription (high)	JPG/PNG
PS Generative Fill / AI	★★☆☆☆	★★★★☆	Low (automatic)	Paid subscription (high) or plugin (medium)	JPG/PNG
GIMP Clone/Heal	★★★☆☆	★★★☆☆	High	Free	JPG/PNG
Photopea (online)	★★☆☆☆	★★★☆☆	Medium	Free (basic)	Multi-format (PSD/PNG, etc.)
PDF editor	★★☆☆☆	★★★★★	Low–medium	Partly free (e.g., WPS)	PDF/Word
OCR + re-typeset	★★☆☆☆	★★☆☆☆	Medium (re-layout)	OCR free/open source; Word, etc.	JPG/PNG/PDF → DOCX
AI image repair	★★☆☆☆	★★★☆☆	Very low	Free/paid (fast tier)	JPG/PNG
Text-to-image generation	★★☆☆☆	★★☆☆☆	Low (generation)	Paid subscription (medium–high)	JPG/PNG
Vectorize + rebuild	★★★★☆	★★★★☆	High	Illustrator paid / Inkscape free	JPG/PNG

Use the matrix as a filter, then use the path section as the procedure. A few readings people get wrong:

Path 1, Content-Aware Fill, is medium difficulty and medium fidelity. It is the simple-background manual tool. It is not weak, and it is not the best you can do on a face.
Path 2, clone and heal, costs the same kind of Photoshop subscription and spends more time to buy a higher fidelity on hard textures. About $20–$50 a month is the subscription band for that seat, trial available. The matrix saying "high" means that band, not a second surprise price.
Path 3, Generative Fill, keeps difficulty low and fidelity high on the rating scale, with automatic time. The version floor is still Photoshop 27.10 or newer. The output is still not pixel-accurate on any given run. The stars describe the ceiling of a good variation.
GIMP and Photopea are the free or free-to-start seats. Their fidelity sits in the middle. Do not move them above clone-and-heal because they are easier to obtain.
The PDF editor's five fidelity stars collapse the moment the page is a scan. WPS on a live text file is allowed to be both inexpensive and excellent. Acrobat being expensive does not raise a scan's rating.
OCR and text-to-image share the low fidelity mark for different reasons: one rebuilds a layout loosely, the other invents a new picture. Neither is a restoration path.
One-click image repair is the fastest row and only a mid fidelity row. Very low time is real. So is the synthetic look on a complex pattern, the paid gate on high-resolution output, and the risk that a free tier watermarks the corner.
Vector rebuild matches clone-and-heal on difficulty, fidelity, and time, with a different cost shape: Illustrator or CorelDRAW paid, Inkscape free. It does not outrank live PDF text.

Path 1 can also open TIFF even though the summary column says JPG/PNG. Read the column as the common delivery formats, not as a ban on other bitmaps Photoshop already opens.

Routing Procedure

Do not pick by habit or by whichever icon is already in your dock. Walk the artifact in this order and take the first branch that fits. If a later check fails, fall through. Do not stack every tool on principle.

The same logic in the order a script would evaluate it:

If the asset is an ID, an invoice, an official document, or a design source file, reject it.
If the PDF probe finds live text and the edit is a string change, take the PDF editor path. This is the five-star case.
If the background is solid or a gentle gradient and you only need the words changed, take the browser replacement first.
If you need a manual desktop pass on that same simple background and you have Photoshop, take Content-Aware Fill.
If the background is photographic or heavily textured and fidelity matters more than the clock, take clone, heal, or patch. If you cannot spend the subscription, take GIMP or Photopea and accept the lower ceiling.
If you have many files, you need speed, and you can reject bad frames, take Generative Fill on Photoshop 27.10 or newer, or a one-click inpainter.
If the layout is mostly regular text and a redesign of the styling is acceptable, take OCR and re-typeset.
If a redesign is not only acceptable but welcome, take text-to-image.
If the art started as vectors and the logo or icons must change, take trace and rebuild.

Reject the input if it is out of scope. IDs, invoices, official documents, and design source files do not enter the router. There is no clever path and no "just this once." A source file that still opens in the original app should be edited there. Flattening it in order to use a bitmap trick is a self-inflicted loss.

Classify the background at 100 percent.

Simple background, solid or a gentle gradient. Use automatic replacement and change the words on a finished image in the browser. That pass is usually under a minute. New users typically get one free edit after signing in, and that edit does not require a credit card. On a desktop, if you already have Photoshop, Content-Aware Fill is the manual equivalent. If the probe found live PDF text, the PDF path is cleaner than either, because you are editing characters instead of inventing pixels.
Complex background: photograph, portrait, dense texture. Do not stay on the one-click path unless you have looked at 100 percent and accepted the scar. Continue to the speed-versus-fidelity choice.

Choose speed or fidelity. You do not get to maximize both.

Speed first, many flyers, routine wording changes. Stay with automatic text replacement, or use Photoshop Generative Fill, or a one-click inpainter. Accept that the AI result is not pixel-accurate. Schedule a human proof of the actual string. The model does not know which digit makes the event land on the wrong weekend.
Fidelity first, high-end print, commercial delivery. Use Photoshop's Clone Stamp or Healing Brush, GIMP, or Photopea. You will sample and paint. Budget about 40 minutes when the texture is genuinely hard, not two minutes with a hopeful filename. About 300 ppi remains the print target while you paint, so you are not matching a thumbnail and then discovering the scar at the printer.

Apply the constraint that actually binds.

No budget, and you need something that runs on your platform. GIMP, or Photopea for a browser session that also opens PSD-style files.
The flyer is a PDF with real text. A PDF editor is the cleanest result when the probe succeeds. Inkscape belongs in this conversation when the text is vector and you want a free editor that also speaks SVG and AI.
Text-heavy, regular layout, and a redesign of the styling is acceptable. OCR and a fresh typeset.
You want creative variants and you are willing to redesign. Text-to-image, with DALL·E under ChatGPT Plus, Midjourney's subscription, Stable Diffusion, or a regional tool such as Wenxin Yige.
The flyer came from a vector export and the logo must change too. Trace and rebuild in Illustrator, CorelDRAW, or Inkscape.

Write the abort line before you export. If the background patch is visible at 100 percent, or the new word's weight is obviously different, you are not done. Switching paths is cheaper than shipping the miss. A second generation on a failed mask is allowed. A third tool that you cannot explain is not a process. Record which branch shipped so the next edit starts from the truth. If the shipped file is a new bitmap, the next edit is a bitmap edit. You did not secretly regain layers.

Five Exit Checks

Whichever branch you ran, the difference between a file that looks fixed and a file that looks like the same flyer is mechanical. Run these five checks in order. They do not require a new product. They require zoom, an Eyedropper, and half a minute of attention you were about to skip.

Match the color, not only the spelling. Sample the original ink with the Eyedropper. Copy a stroke or a shadow if the old word had one. The correct font in the wrong color is obvious at a glance. If the old word was knocked out of a photo and the new word is solid black, you skipped this check. Sample again from the old file, not from your repaired patch, or you will sample your own mistake and lock it in.
Keep the resolution consistent with the destination. About 300 ppi for print. If you typeset at screen resolution and drop the word onto a print file, the new type looks softer or sharper than the sheet around it. Match the source to the output before you hunt for a better font. The font is often innocent. Also match sharpness the other way: a razor-sharp vector word dropped on a soft JPEG looks pasted on, even when the family is right. A tiny blur or the original's texture, used sparingly, is a match technique, not a defect.
Set weight, then spacing. Weight is the usual reason replacement type looks wrong. Fix weight first. Then tighten or loosen letter spacing, especially on headlines and logo lines that were tracked tight. Do not start by changing the font family if the family is already close. You will wander through a font menu and still miss the tracking. For a price, also check the figure style. A lining figure next to old-style figures, or a much wider currency sign, can make a correct font look wrong.
Feather and blend the repair. A feathered selection and a low-opacity cleanup pass make Content-Aware fills and heals much harder to see. A hard rectangular patch shaped like the old bounding box fails this check even when the colors are close. Shift+F6 is the feather shortcut in Photoshop. In the other editors, use whatever feather or expand control you have, then a soft cleanup. The principle is the same. On GIMP, the Blur tool on the fringe is the cousin of that feather. On a PDF text edit you usually skip this check, because you never made a hole. If you did make a hole, you are no longer in the PDF-text success case.
Review at full resolution for about 30 seconds before export. Artifacts and small font mismatches hide in a thumbnail and shout at 1:1. Thirty seconds here is the cheap catch. Finding the same miss after the flyer is distributed is a different, worse ticket. Look specifically at counters inside letters, diacritics, the shadow side of the patch, any edge where the old word met a photo, and the characters of the date or price. Fit-to-screen is not this review. If you cannot zoom to 100 percent in the tool you used, download and zoom somewhere else before you send the file.

If check 5 fails, return to the path you were on. Do not export and fix it in a chat thread. Name the signal, using the list above, and either repair inside the same path or switch. Shipping a known halo because the meeting started is how the flyer gets reprinted.

Which Branch to Run First

Losing the PSD or the Canva link does not mean you must rebuild the flyer from a blank page. It means the input is a finished picture, or a PDF that might still be text, and you owe it a branch choice based on what a probe actually shows.

Run the probes first. Text cursor on a PDF. One hundred percent zoom on the background behind the target words. A written edit intent that says remove or replace, and that records whether the new string is longer. A scope check that throws out IDs, invoices, official documents, and design source files. Those four actions take less time than opening the wrong desktop app and painting yourself into a hole.

If the background is clean or simple, which is true of most flyers, run the automatic replacement first. Upload, change the words, download. On a straightforward file that is usually under a minute, and the first edit after sign-in typically does not ask for a card. If that 100 percent view is clean, stop spending. If the background is genuinely hard and the sheet has to survive a print inspection, the sampling tools still earn the subscription. Clone and heal is the high-fidelity, high-time branch for that texture, and the clock for a real salvage is about 40 minutes, not a coffee break. Generative Fill can sit between those poles when you have Photoshop 27.10 or newer and you will reject bad variations instead of shipping the first one. Remember that those variations are not pixel-accurate. Live PDF text outranks all of the pixel repairs, and only then does the five-star rating apply. A scan does not borrow it.

Use the other branches for the constraints they actually answer. GIMP when the budget is zero and you can paint. Photopea when you need a browser and PSD-style files more than you need Photoshop's fill. OCR when the flyer is a regular, text-heavy layout and a cousin of the design is good enough. One-click inpainting when removal speed matters and you will still typeset afterward, watching for a free-tier watermark and a resolution cap. Text-to-image when a redesign is the point, with the logo and the QR pasted back from real assets. Vector trace when the mark itself has to change and the source was line art, knowing the trace can fail into jagged nodes.

Take the flyer that is stuck in the queue. Replace the copy on a flattened flyer and see how close the one-minute pass gets before you commit to the 40-minute Photoshop salvage. If the full-resolution view is clean, ship it and keep the original beside the export. If it is not, name the failure signal in front of you and switch to the branch that repairs that signal. You still do not need the missing design file to make that choice. You need the input class, the background class, and a refusal to ship the scar you can already see.

Treat Burned-In Text as a Build Artifact: A 2026 Developer Workflow Without Photoshop

baiwei — Thu, 08 Oct 2026 10:52:53 +0000

A finished photo is a compiled artifact. The letters are not a text layer you can retype. They are pixels blended into a wall, a label, a face, or a gradient. If your first move is still “open Photoshop and learn the clone stamp,” you are debugging the binary with a tutorial that can take hours to absorb, then spending dozens of minutes on a single file.

I hit that wall often enough that I stopped collecting tips and started treating removal as a routing problem. Over about a week I ran hundreds of images through 10 non-Photoshop tools: flat color, product shots, ID-style portraits, indoor rooms, outdoor scenes, fine grids, and backgrounds dense enough that a healing brush has to guess. The result was not a universal winner. No single tool handles every background. Simple fields and complex ones — faces, gradients, dense detail — need different logic. AI repair already covers about 80% of daily needs. The other 20% is where sampling, layers, and a slow hand still earn their keep.

What actually changed my speed was refusal. Matched tools compress work that used to take dozens of minutes down to a few seconds. Unmatched tools smear texture or trap you in pixel nudging. The rest of this piece is the workflow I now use: classify the background, decide whether a human should touch the pixels at all, then call one operator with the shortcuts and limits that operator actually has.

Guard clauses: classify the background before you touch a brush

Think of the image as an input, not as a canvas you are morally obliged to paint. The first function in the pipeline is a classifier, and it only has to answer one question: is the area behind the type predictable?

A predictable background is a solid color, a simple gradient, or a repeating texture a fill can copy from next door. An unpredictable background includes a face, architectural detail, uneven light, or any patch where neighboring pixels do not agree with each other. Using a heavy manual editor on the first kind wastes the afternoon. Using a one-click eraser on the second kind leaves a smudge that you will notice in review, even if a thumbnail looks fine.

That split is the whole method. I do not open a healing brush because a file “looks important.” I open it because the background failed the simple-case test and the delivery needs a careful sample. For everything else I want a narrower program: a batch eraser, a phone heal for a corner date, or a text-specific editor that boxes the existing line and writes a replacement.

Three refusals keep me out of manual mode:

The mark is small, the field behind it is simple, and I am on a phone. A finger-sized heal is enough. A desktop layer stack is ceremony.
I have a pile of stock frames and the only defect is a watermark on a flat or lightly textured field. Sampling each one by hand does not scale.
The job is to change or clear type that is already printed on a label, badge, poster, menu, or cover. I do not need a general paint program, and I should not be anywhere near an ID card, an invoice, an official document, or someone’s design source file.

Manual retouching stays on the table for portraits, fussy commercial posters, and any complex texture where I can point at a clean neighbor and trust the blend. That work has a high ceiling. It is also the wrong default. If I cannot explain, in one sentence, why the automated path would fail this file, I do not get to spend several minutes stroking pixels.

The manual runtime: layers, samples, and three editors

When the guard clause fails, the implementation is old and still correct. You protect the original, you read a clean pixel, you write over the glyph, and you let a blend mode or a heal kernel hide the seam. The keyboard is the API. The mouse is just how you aim it.

I keep three editors in this branch. They are not substitutes for each other, even though all three can end with new type on top of a repaired patch.

Photopea

Photopea is the free, no-install editor I open when I want Photoshop-shaped control inside a browser tab. There is no registration. It runs on the site, it opens a normal image, and it also opens a PSD, which matters when someone hands you a half-finished file instead of a flat export.

How I run it

Open the Photopea site. Nothing to install, and no account is required before the canvas loads.
Use File → Open and import the image or the PSD.
If the background is simple, press B for the paint brush. Hold Alt, click a clean patch of the background to sample that color, then paint straight across the letters.
If the background is complex, press J for the heal brush. Hold Alt and sample close to the text, not from a random corner of the frame. Paint over the glyphs. The heal pass blends the edges instead of dropping a flat color onto texture.
Press T, set the new wording, adjust layer styles until the type sits in the same light as the photo, then use Export As.

What it costs in time, and what the pixels look like

A careful pass takes several minutes. That is the point. Done properly, the repair reaches a professional seamless level that the one-click tools do not promise. Done carelessly — sampling from the wrong side of a highlight, or painting a hard brush across a gradient — it looks like every other bad clone job. The ceiling is high only if you treat Alt as a read and the brush as a short write.

What I keep it for, and where it breaks

I keep Photopea because the feature set is deep and PSD files survive the round trip. I drop it when the network is unreliable, because the whole runtime lives in the browser, and when the person doing the job has not learned the shortcuts yet. The learning curve is steeper than an upload-and-download eraser. I also do not use it as the first stop for a simple wall of color. Several minutes of sampling is a bug if a few-second tool would have filled the same patch.

One implementation habit: decide simple versus complex before you press B or J. The paint brush is a solid-color write. The heal brush is a neighborhood blend. Mixing them up is how flat walls get textured noise and how detailed walls get muddy discs.

GIMP

GIMP is the open-source desktop option for the same class of job, aimed at someone who already knows what a sample source is and wants authority over every pixel. I use it when the file should not depend on a website staying up, and when I expect the repair to take long enough that a local undo stack matters.

How I run it

Install GIMP if it is not already on the machine, open it, and import the image.
Create a new layer before any repair. That layer is the patch. The imported image stays untouched underneath, which is the only sane way to keep a manual session reversible.
For the Clone tool, hold Ctrl and click a clean area with no text. That click is the sampling source. Then paint only on the glyphs. If the source and the destination drift apart in brightness, stop and sample again instead of scrubbing harder.
For a gradient or a more complex texture, switch to the Heal tool. Sample the same way and paint over the text. GIMP matches color as it blends, which is what you want when the background is not one RGB value.
Move to the Text tool, set the replacement copy, add layer styles such as a shadow if the original type had depth, and export.

What it costs in time, and what the pixels look like

Budget several minutes to dozens of minutes. There is no batch shortcut hiding in this path. In exchange, background detail can come out smooth and natural, and the ceiling is as high as your sampling discipline. A new source click every time the texture changes will beat one heroic brushstroke every time.

What I keep it for, and where it breaks

GIMP is completely free, and the control is 100% manual. That second fact is also the drawback. The learning curve is steep. I do not put a teammate who has never sampled a clone source in front of GIMP on a deadline. I also do not use it to wipe a date stamp off a corner of sky. The tool will do it. The time accounting will not survive.

The layer rule is non-negotiable in my own files. Repairing on the base layer feels faster for the first minute and becomes expensive the moment a stroke crosses from the letter into a highlight you needed to keep. A separate layer is a commit you can hide, mask, or throw away without reopening the original export.

Pixlr

Pixlr sits between the two manual editors and the upload-and-forget AI tools. It ships as a professional editor, Pixlr E, and a simpler editor, Pixlr X. I open it when I want layers and a selection, but I do not want to install a desktop suite, and I am willing to live with whatever the free session puts on screen.

How I run it

Open the Pixlr site and load the image. No registration is required to start.
If the type is high contrast against the background, select it with the magic wand or the lasso, then run content-aware fill from the edit tools. The selection is doing the work a brush would do. A tight selection matters more than a second pass.
If the selection route is awkward, switch to AI Object Remove and paint across the text so the reconstruction can run on that stroke.
Add the new wording with the text tool and save from the browser session.

What it costs in time, and what the pixels look like

Wall-clock time follows the network. On simple scenes the fill behaves well, and the stated output clarity goes up to 1200 dpi. I do not treat that number as a promise that a busy background will reconstruct cleanly. It is a ceiling on the file you can get out, not a description of every inpaint.

What I keep it for, and where it breaks

Layers are available, and you can work without creating an account. The free session carries a lot of ads. AI removal may require a premium plan, so I do not depend on the Object Remove path for a workflow I have to repeat on a free seat. When the type sits on a face or a messy texture, I would rather sample in Photopea or GIMP than hope a wand selection understood the edge.

Pixlr is the editor I recommend when someone asks for “Photoshop in a tab” and then admits they only need a selection plus a fill. It is not the editor I leave open for a fifty-image queue.

The batch runtime: call an eraser, then review the seam

Most daily files never deserve the manual runtime. Once I know the background is ordinary — a product sweep, a room, a landscape, a reasonably even surface — I want an algorithm to find the ink and rebuild the patch. That is the 80%. My job becomes review: zoom to 100%, look at edges, and only then decide whether a human sample is required.

Three tools live on this path. They feel similar from the outside and fail in different places.

LightPDF

LightPDF is the browser tool I use when the priority is throughput and the scene is a natural landscape or an indoor shot. The backend model detects type and reconstructs the background. On those scenes the fill has been strong enough that I stopped assuming “automatic” meant “blurry.”

How I run it

Open the image text-removal page on the LightPDF site in a browser.
Upload a JPG, a PNG, a HEIC, or another format the page accepts.
Let it detect and remove the text on its own, or mark a specific area if you only want one line gone.
Wait a few seconds while the reconstruction finishes, then download the result.

What it costs in time, and what the pixels look like

The pass is a few seconds, not a coffee break. The file I get back keeps the original resolution, and detail on ordinary photographs holds up. I still inspect fine structure before I ship the frame. Speed does not excuse a skipped review.

What I keep it for, and where it breaks

The operation is about as short as a consumer web tool gets, and batch processing covers up to 50 images at once. That is the feature that retires a weekend of clone-stamping. The free tier limits how many high-quality runs you can take, so a large queue can stall even when the model itself is fast. Extremely fine grids are the visual failure I watch for: the reconstruction occasionally breaks the pattern. If the background is a mesh, a screen, or a tight repeating lattice, I do not trust the first download without a zoomed check, and I am ready to move that one file to a sampling editor.

I also do not send LightPDF a job whose real goal is “set this label to a new sentence in a similar style.” It is an eraser and a reconstructor. Replacement copy is a later step in some other tool.

Meitu

Meitu, on PC or on the web, is the AI editor I reach for when the lettering is Chinese, printed or handwritten, and the frame is the kind of product photo or ID photo that shows up in domestic social posts. The recognition work is where it has a real edge. I am not using it as a general Western-layout compositor.

How I run it

Open the Meitu desktop app or the web design tool.
Import the image and go into the seamless text-modification mode inside AI image editing.
Select the characters or symbols that should disappear. The model reads the surrounding texture and fills it.
If a stroke or a shadow remains, switch to the AI smart-erase tool in local image editing and clean that residue by hand. This is a small correction, not a full manual repaint.
Use the text tool to set new type, then export.

What it costs in time, and what the pixels look like

A single removal is about 3 seconds. On product images and ID photos the fill has been clean, without an obvious scar where the old line sat. That is the quality bar I hold it to. If I can see a patch in the first look, I do not talk myself into shipping it.

What I keep it for, and where it breaks

Chinese text handling is the reason it stays installed, and the same session can add the replacement line, so removal and retypesetting are one path instead of an export to a second app. Some advanced features need a VIP membership. Very complex backgrounds are a limited case: the three-second fill is not a substitute for a sampled heal when the texture behind the type is fighting the model. I keep Meitu on the speed branch of the router, next to LightPDF, and I take it off that branch as soon as the background stops being regular.

There is no screenshot in my notes for Meitu. I judge it from the exported file, not from a marketing page.

PhotoWorks

PhotoWorks is desktop software built to feel foolproof: repair and watermark removal sit under a small interface, with a stack of photo filters behind them if the frame also needs ordinary finishing. I use it when the person running the job wants one installed app and does not want to learn brush shortcuts.

How I run it

Install PhotoWorks, launch it, and load the photo.
Choose the repair tool or the watermark-removal tool.
Mark the text yourself, or turn on automatic recognition and let the software replace that area in one action.
If the frame needs new wording, use the built-in editor to add it, then export a high-quality file.

What it costs in time, and what the pixels look like

The edit finishes within about a dozen seconds. Clarity and the general natural look usually hold. “Usually” is doing real work in that sentence. I have seen the AI repair come back slightly smudged, which is acceptable for a casual post and not acceptable for a file someone will print large.

What I keep it for, and where it breaks

The filter set makes it a reasonable place to finish a photograph after the type is gone, so removal and light grading do not require a second install. The software is paid. I do not put it on a machine just to erase one date. I also do not trust the one-click replace on a background where a smudge would show, such as skin or a sharp repeating pattern. When that risk is real, a dozen seconds saved is not worth the retake. Photopea’s J shortcut or GIMP’s Heal tool is the slower, more honest path.

PhotoWorks belongs with the other AI erasers because the human is mostly marking a region. It does not belong with GIMP. The minimal UI hides the algorithm; it does not give you a sample source.

Narrow binaries: one defect, one program

A general editor is the wrong shape for some jobs. Two tools in this set do not try to be Photoshop. That restraint is the feature.

Inpaint

Inpaint is a desktop eraser I keep for stock frames and other batches where the only task is to wipe a mark off a solid or textured field. It does not try to be a layout program. You select, you erase, you nudge with a healing brush if a trace remains.

How I run it

Install the Inpaint client.
Select the text with the rectangle or the lasso. Include the full glyph and the immediate fringe, not a loose region of innocent background.
Hit Erase. The program fills the selection from the surrounding texture immediately.
If a halo or a repeated chunk of texture remains, use the healing brush on that residue only, then save.

What it costs in time, and what the pixels look like

Figure a few seconds per image. On solid color and on ordinary texture the fill is as close to ideal as this kind of tool gets. The selection still matters. A rectangle that slices through a shadow will rebuild the shadow wrong, and no amount of “it’s only a watermark tool” fixes a bad mask.

What I keep it for, and where it breaks

The focus is the advantage. Batch watermark cleanup is the job I assign it, because I am not paying the overhead of a full editor for a mark on an otherwise usable stock photo. The same focus is the limit. Inpaint does not add new text by itself. If the next requirement is a replacement headline, I export and finish the type somewhere else. I also do not expect it to invent a plausible face behind a caption. Solid and textured backgrounds are its range. Outside that range I am back to a sampled heal.

Treat Inpaint as a filter in a directory script, not as a design suite. Input frame, mask the words, write the cleaned file, stop.

ReWords AI

ReWords AI is the narrow web tool I use when the artifact is already finished and the task is the type itself. It fits labels, badges, posters, menus, and covers: either clear the line or set a new one in that same printed spot. People who need to remove text from images and then put different wording in the same place are the audience. It is not a tool for ID cards, invoices, official documents, or design source files. If the job is any of those, I stop. Editing the bitmap is the wrong operation, and this product is the wrong program.

How I run it

Open the page in the browser and upload a finished image. PNG, JPG, and WebP are the formats I actually feed it.
Box-select the existing text. The box is the whole contract. Sloppy bounds pull in neighboring print or miss a descender, and the fill has to guess at a region you defined badly.
Leave the replacement field blank if the line should simply disappear. Type the new wording if the line should change.
Sign in. Generation happens after login. Then download the image the page returns.

What it costs in time, and what the pixels look like

The session stays in the browser. Fonts, colors, and the surrounding background stay as close to the original as the model can manage. The fill is best-effort. On a complex background I do not expect a seamless result, and I do not tell anyone else to expect one. Close and seamless are different outcomes. I zoom the boundary between the new patch and the original print before I accept the file.

What I keep it for, and where it breaks

Nothing has to be installed, and a free trial is available without a credit card. The trial quota is limited, so a long queue still runs out of tries. It will not recover the original font file; you get pixels that resemble the source, not an editable font you can drop into a design app. On complex backgrounds it only does its best. That is an honest limit, and it is why a commercial poster with a difficult texture still goes to Photopea or GIMP when the seam has to disappear.

I also keep the scope tight on purpose. A label, a badge, a poster, a menu, or a cover is in range because the type is already part of a finished picture. A source file that still has live text objects should be edited in the authoring tool, not flattened and repainted. An identity document or an invoice should not be in the pipeline at all.

The developer habit that matters here is the blank field. Removal and replacement are the same control. Empty means erase. A string means change. I do not look for a separate “delete text” mode and then a second product to typeset, unless the background complexity has already kicked the file out to a manual editor.

Edge devices: the phone is a workstation only for small jobs

Leaving the desk does not mean the pipeline disappears. It means the input device changes and the acceptable defect size shrinks. A phone is a good heal for a corner mark and a good layout toy for a social cover. It is a bad place to sample a commercial face at full resolution.

PicsArt

PicsArt is the phone app I use when the output is a short-video cover or a social image and I want object removal plus a loud font, not a forensic rebuild. The asset library is the product as much as the eraser.

How I run it

Open the image on the phone, go to the tools tab, and choose text inside object removal.
Brush around the words, or let the app detect them, then run remove.
Open the text tools, pick an effect font, and set the new title.
Export to the camera roll or share from the app.

What it costs in time, and what the pixels look like

The automatic reconstruction keeps a natural read of the original texture and color. Large images run a bit slower than a small cover. If I am processing a full-resolution photo just to change a caption, I would rather move that file to a desktop eraser than wait on the handset and then discover the seam at export size.

What I keep it for, and where it breaks

The font and sticker library is large, and the gestures are obvious, which is why it wins for a poster remade on a phone. Some of the more advanced AI models sit behind a subscription. I do not plan a production step around a model I might not have on a free account. PicsArt is the mobile remake path in the router. It is not the path for a batch of 50, and it is not the path when I need desktop sampling.

Snapseed

Snapseed is the free Google editor I use for the smallest version of this problem: a date in a corner, or a small watermark, on a simple background. There is no ad layer in the way. There is also no ambition to rebuild a busy scene.

How I run it

Import the photo on the phone and choose healing from the bottom toolbar.
Paint over the text with a finger. Short strokes. The tool erases by pulling in the nearby field, so a small target on a calm background is the whole design.
If the frame needs new wording, open the text tool and apply one of the artistic templates.

What it costs in time, and what the pixels look like

It is a quick swipe. On a small area with a simple background the result is genuinely good, which is why I keep the app on the phone at all. The moment the area grows, or the background stops being simple, the same swipe piles into a smudge. Overworking the spot makes it worse, not better. I stop at the first clean pass or I abandon the file to a desktop tool.

What I keep it for, and where it breaks

Completely free and free of ads is a rare combination, and for a corner date it is the right amount of software. It does not support a serious removal across a large region or a complex background. I do not use it on a face, a storefront full of detail, or a poster that has to be retyped precisely. Those files feel editable because the heal gesture is always available. They are not editable well. Knowing that is the difference between a ten-second fix and a file I have to reshoot or rebuild.

Snapseed is the mobile exception inside the simple-background branch. If I am at a desk, I usually will not bother. If I am holding the phone and the defect is tiny, I will not wait until I can sit down.

The routing function I actually run

Intuition is a bad scheduler. After the week of tests, the dispatch I follow is short, and I keep it in this order so I do not “just try Photoshop” out of habit.

if background is solid or a simple gradient:
    if the edit is small and I am on a phone:
        Snapseed healing, then stop
    else:
        ReWords AI
        blank replacement -> remove
        typed replacement -> new wording
elif background is complex (portrait, architecture, dense texture):
    if the queue is large or the need is daily speed:
        LightPDF on the web, or Meitu on PC
    elif the delivery needs a careful commercial finish:
        Photopea (J + Alt) or GIMP (Heal, new layer, Ctrl sample)
elif the job is a batch of watermarks on stock frames:
    Inpaint, then add type elsewhere if type is required
elif the job is a poster remade on a phone:
    PicsArt

Read that as policy, not as a scoreboard. A simple background goes to ReWords AI when I need to edit text in images: leave the replacement blank to remove the line, or type the new line and generate. The same simple case, if it is only a small mobile touch-up, still goes to Snapseed. I do not promote a simple file into Photopea just because Photopea can do it.

A complex background does not skip the second question. Speed, large batches, and ordinary daily images go to LightPDF in the browser or to Meitu’s AI on the PC, and the algorithm does the fill. Ultimate quality — a high-end poster, a commercial file where a smudge would be a rejection — goes to Photopea or GIMP, and I sample with the heal brush by hand. Architecture and portraits are the usual reasons I take that slower branch. A fine grid that already failed inside LightPDF can take it too, even if the rest of the scene looked easy.

Batch watermarks stay with Inpaint. A poster I am remaking away from the desk stays with PicsArt. PhotoWorks remains available inside the “dozen seconds, accept a possible smudge, filters afterward” niche, not as the quality path and not as the fifty-image path. Pixlr remains the in-between browser editor when I want a wand, a lasso, or content-aware fill and I can tolerate ads, with AI removal treated as optional rather than guaranteed on a free plan.

Two sentences keep the policy honest. AI repair is enough for about 80% of what shows up in a normal week, which is why manual sampling is a branch and not a lifestyle. And the tools that feel primitive — a heal swipe, a rectangle and an Erase button, a box and a replacement string — are faster than a famous paint program precisely because they refuse every job that is not theirs.

If you want the difference on your own files, do not start with the easiest thumbnail on the desktop. Pull the frame with the most complicated background, open Photopea or the LightPDF web tool, and run that one image through the branch the classifier picked. The shortcut, or the decision not to use a shortcut, shows up immediately. The hours you might have spent learning a clone stamp for a flat wall do not.

SeedAudio: natural AI voice audio from scripts in your browser

baiwei — Mon, 28 Sep 2026 23:20:31 +0000

Designed for content creators and production teams, SeedAudio turns written scripts into natural AI voice audio inside your internet browser. Users have access to comprehensive delivery controls to modify pace, pitch, and vocal dynamics before finalizing tracks.

The platform emphasizes permission-aware voice workflows, ensuring that voice creation complies with authorizations. Direct audio downloads are supported for finished tracks, and a freemium plan keeps core workflows accessible.

Learn more at https://seedaud.io/ — SeedAudio focuses on authorized script-to-speech with natural voice output for everyday production work.

Introducing iptv-org pro: A Public TV Channel Directory

baiwei — Tue, 01 Sep 2026 17:00:59 +0000

iptv-org pro is an independent web directory indexing public television channels sourced from the open-source iptv-org dataset. Functioning as a public TV channel directory, it provides a centralized platform to browse publicly available television information.

Each entry brings together official station websites, language tags, category classifications, and listed public stream endpoints into a single searchable record. This structure allows users to explore and look up channel details, classifications, and public endpoints in one place.

You can explore the directory at iptv-org pro.

AI Text Watermark Remover: Detect Invisible Characters and Clean AI Residue

baiwei — Tue, 01 Sep 2026 16:59:49 +0000

AI Text Watermark Remover is a free tool designed to detect invisible characters, strip Markdown residue, and regenerate drafts when AI models embed watermarks into the wording itself. It supports text generated by systems such as Claude, ChatGPT, Gemini, and more.

The invisible-character detection and Markdown cleaning run locally in your browser, and the tool does not store text entered into the free version. Its core focus is to clean hidden marks while keeping your meaning intact.

You can try the tool at AI Text Watermark Remover.