<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Saihhold Zhao</title>
    <description>The latest articles on DEV Community by Saihhold Zhao (@saihhold_zhao_064664f95f9).</description>
    <link>https://dev.to/saihhold_zhao_064664f95f9</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F1833208%2F5916e4a5-1831-4c51-ac39-c6d329430228.jpg</url>
      <title>DEV Community: Saihhold Zhao</title>
      <link>https://dev.to/saihhold_zhao_064664f95f9</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/saihhold_zhao_064664f95f9"/>
    <language>en</language>
    <item>
      <title>How to Edit Text in an Image with AI (and Keep the Original Font)</title>
      <dc:creator>Saihhold Zhao</dc:creator>
      <pubDate>Sat, 24 Oct 2026 05:45:39 +0000</pubDate>
      <link>https://dev.to/saihhold_zhao_064664f95f9/how-to-edit-text-in-an-image-with-ai-and-keep-the-original-font-oo4</link>
      <guid>https://dev.to/saihhold_zhao_064664f95f9/how-to-edit-text-in-an-image-with-ai-and-keep-the-original-font-oo4</guid>
      <description>&lt;p&gt;You have a flat image — a screenshot, a poster, a photographed menu — and one word in it is wrong. No PSD, no layers, no idea what font it uses. The classic fix is opening Photoshop, cloning out the old text, guessing the font with Match Font, and nudging kerning until it looks right. For a one-word change, that's a lot.&lt;/p&gt;

&lt;p&gt;This post walks through the faster workflow I now use: AI-based in-place text replacement. I'll use &lt;a href="https://retypit.ai/" rel="noopener noreferrer"&gt;Retypit&lt;/a&gt;, a browser tool built specifically to &lt;a href="https://retypit.ai/" rel="noopener noreferrer"&gt;edit text in an image&lt;/a&gt; while keeping the same font, color and style.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why this is harder than it looks
&lt;/h2&gt;

&lt;p&gt;Changing text inside a raster image is really three problems:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Finding the text&lt;/strong&gt; — the image is just pixels; there's no text layer.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Removing the old words&lt;/strong&gt; without leaving a smudge on the background, texture, or gradient behind them.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Re-rendering the new words&lt;/strong&gt; so they match the original typeface, weight, color, shadow and wear.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;General-purpose generative fill tends to fail on step 3: it regenerates a whole region and the letters come back in a generic sans-serif, or slightly garbled. A dedicated tool handles all three steps together and touches only the line you changed.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 1: Upload the image
&lt;/h2&gt;

&lt;p&gt;Go to &lt;a href="https://retypit.ai/#upload-zone" rel="noopener noreferrer"&gt;retypit.ai&lt;/a&gt; and drop in a PNG, JPG or WebP (up to 20 MB). You'll need to sign in (Google or email). The tool detects every line of text automatically and puts a marker on each one, with an editable list beside the image.&lt;/p&gt;

&lt;p&gt;Tip: use the highest-resolution version you have. Detection and re-rendering both work better with crisp letterforms.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 2: Click the line and retype it
&lt;/h2&gt;

&lt;p&gt;Select the line you want to change and type the replacement. Some examples from my own use:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;A flyer: &lt;code&gt;SUMMER&lt;/code&gt; → &lt;code&gt;WINTER&lt;/code&gt;, &lt;code&gt;SAT JUL 12 · 7PM&lt;/code&gt; → &lt;code&gt;SUN JAN 18 · 4PM&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;A menu photo: &lt;code&gt;$14&lt;/code&gt; → &lt;code&gt;$19&lt;/code&gt; across several prices&lt;/li&gt;
&lt;li&gt;A UI screenshot: fixing a button label for docs without rebuilding the UI&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;You can change several lines in one go, then press &lt;strong&gt;Edit text&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 3: Download
&lt;/h2&gt;

&lt;p&gt;The AI rebuilds the new words in the original lettering — font, color and texture — and leaves faces, backgrounds and layout alone. That "only the text changes" behavior is the main reason I use it as an &lt;a href="https://retypit.ai/#features" rel="noopener noreferrer"&gt;AI image text editor&lt;/a&gt; instead of a generic inpainting model. Download the result and you're done.&lt;/p&gt;

&lt;h2&gt;
  
  
  Use cases where this shines
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Screenshots for docs and tutorials.&lt;/strong&gt; If you write technical docs, you know the pain of a typo in a screenshot of a UI you can't easily reproduce. You can &lt;a href="https://retypit.ai/#use-cases" rel="noopener noreferrer"&gt;edit text in a screenshot&lt;/a&gt; instead of re-capturing it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Fixing typos in AI-generated images.&lt;/strong&gt; Image models still misspell words. Correcting one word beats rerolling the whole scene and losing the composition you liked.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Posters, flyers and menus.&lt;/strong&gt; Reusing last month's event poster with a new date, or updating prices on a photographed menu.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Product photos.&lt;/strong&gt; Swapping label text on packaging mockups.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Memes and social graphics.&lt;/strong&gt; Retyping captions while keeping the original look.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  When to still use Photoshop
&lt;/h2&gt;

&lt;p&gt;To be fair: this is not a Photoshop replacement. If you need full compositing, layout changes, or you actually have the layered source file, use your design tool. And it's not a font identifier — if you need to know the font's &lt;em&gt;name&lt;/em&gt;, use a font-ID tool. The sweet spot is: &lt;strong&gt;you only have the flat image, and you just need to &lt;a href="https://retypit.ai/#how-it-works" rel="noopener noreferrer"&gt;change text in a photo&lt;/a&gt; so it looks like it was always there.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Also worth knowing: the tool aims to change only the selected text, but minor changes elsewhere can occasionally happen, so zoom in and check edges before you ship the image.&lt;/p&gt;

&lt;h2&gt;
  
  
  Quick comparison
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Approach&lt;/th&gt;
&lt;th&gt;Font match&lt;/th&gt;
&lt;th&gt;Background preserved&lt;/th&gt;
&lt;th&gt;Time for 1 word&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Manual Photoshop (clone + Match Font)&lt;/td&gt;
&lt;td&gt;Depends on your skill&lt;/td&gt;
&lt;td&gt;Yes, if careful&lt;/td&gt;
&lt;td&gt;10–30 min&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Generic AI inpainting&lt;/td&gt;
&lt;td&gt;Often generic font&lt;/td&gt;
&lt;td&gt;Region is regenerated&lt;/td&gt;
&lt;td&gt;1–2 min, many retries&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Dedicated in-place text editor&lt;/td&gt;
&lt;td&gt;Matches original style&lt;/td&gt;
&lt;td&gt;Only the line changes&lt;/td&gt;
&lt;td&gt;Under a minute&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Wrapping up
&lt;/h2&gt;

&lt;p&gt;For one-off text fixes in flat images, a purpose-built tool saves a surprising amount of time. If you want to try it, &lt;a href="https://retypit.ai/" rel="noopener noreferrer"&gt;Retypit&lt;/a&gt; lets you &lt;a href="https://retypit.ai/#compare" rel="noopener noreferrer"&gt;replace text in an image without Photoshop&lt;/a&gt; right in the browser. I'd love to hear in the comments what kinds of images you most often need to fix — screenshots? AI art? menus?&lt;/p&gt;

</description>
      <category>ai</category>
      <category>design</category>
      <category>tutorial</category>
      <category>productivity</category>
    </item>
    <item>
      <title>Annotation &amp; Comment: A Simpler Approach to AI Image Editing</title>
      <dc:creator>Saihhold Zhao</dc:creator>
      <pubDate>Wed, 23 Sep 2026 05:28:13 +0000</pubDate>
      <link>https://dev.to/saihhold_zhao_064664f95f9/annotation-comment-a-simpler-approach-to-ai-image-editing-1058</link>
      <guid>https://dev.to/saihhold_zhao_064664f95f9/annotation-comment-a-simpler-approach-to-ai-image-editing-1058</guid>
      <description>&lt;h2&gt;
  
  
  Introduction
&lt;/h2&gt;

&lt;p&gt;Most AI image editing tools today rely on &lt;strong&gt;text-only prompts&lt;/strong&gt; to describe what needs to be changed and how it should be modified.&lt;/p&gt;

&lt;p&gt;However, when an editing task becomes more complex, writing prompts can quickly become cumbersome. More importantly, text alone is sometimes not precise enough to indicate the exact area of an image that needs to be edited.&lt;/p&gt;

&lt;p&gt;For example, suppose I want to make several changes to the image below:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Make the model hold a specific object&lt;/li&gt;
&lt;li&gt;Change the color of her shoes to red&lt;/li&gt;
&lt;li&gt;Replace her outfit with a specific piece of clothing&lt;/li&gt;
&lt;li&gt;Change the background environment to a café&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;With the traditional text-only approach, I might need to write a prompt like this:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Make the model raise her hand and hold the coffee cup from Image 1, change her shoes to red, replace her outfit with the clothing from Image 2, and change the background to a café.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwwajg8lnatxtl97jpbwa.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwwajg8lnatxtl97jpbwa.png" alt=" " width="799" height="502"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Annotation &amp;amp; Comment Editing
&lt;/h2&gt;

&lt;p&gt;I simplified this process with a new interaction method.&lt;/p&gt;

&lt;p&gt;Instead of writing a long prompt, users only need to &lt;strong&gt;place markers directly on the image and briefly describe the desired change in the corresponding input fields&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;This makes it possible to tell the model exactly &lt;strong&gt;where&lt;/strong&gt; a change should happen through visual annotations, while using short text instructions to describe &lt;strong&gt;what&lt;/strong&gt; should be changed.&lt;/p&gt;

&lt;h2&gt;
  
  
  Implementation and Real-World Testing
&lt;/h2&gt;

&lt;h3&gt;
  
  
  1. Choose an Image Editing Model
&lt;/h3&gt;

&lt;p&gt;So far, this method has worked well with the following image editing models:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;GPT Images 2.5 Flare / Sunburst&lt;/li&gt;
&lt;li&gt;Nano Banana 2&lt;/li&gt;
&lt;li&gt;Seedream 5.0 Pro&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Based on these tests, I believe the same approach is also worth testing with newer generations of image editing models released after them.&lt;/p&gt;

&lt;p&gt;The following example demonstrates an actual test using &lt;strong&gt;GPT Images 2.5 Sunburst&lt;/strong&gt; in &lt;strong&gt;PoloX AI&lt;/strong&gt; &lt;a href="https://polox.ai" rel="noopener noreferrer"&gt;PoloX AI&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;The implementation is open source and available on GitHub:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/saihhold-zhao/polox_ai" rel="noopener noreferrer"&gt;https://github.com/saihhold-zhao/polox_ai&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Use the Image Annotation Edit Skill
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;1. Trigger the Skill&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Use:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;/annotated-image-edit&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;to trigger the &lt;strong&gt;Image Annotation Edit Skill&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7j5g3z3rb1lgszbppr0l.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F7j5g3z3rb1lgszbppr0l.png" alt=" " width="799" height="346"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;2. Enter Annotation Editing Mode&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Select &lt;strong&gt;Annotation Editing Mode&lt;/strong&gt; to open the image annotation tool.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;3. Add Markers and Editing Instructions&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Place a marker directly on the part of the image you want to modify, then enter a short instruction in the corresponding input field.&lt;/p&gt;

&lt;p&gt;For example, if you want to change the shoes to red:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Place a marker on the shoes → Enter “red”&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;That's it.&lt;/p&gt;

&lt;p&gt;If you want to use a reference image, you can upload it directly in the corresponding input field.&lt;/p&gt;

&lt;p&gt;For example, if you want the model to hold a specific coffee cup:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Place a marker on her hand → Enter “raise her hand and hold this” → Upload the coffee cup reference image&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;This creates a clear relationship between the target location, the editing instruction, and the reference image.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4a93vmqfk8i94emxbe3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4a93vmqfk8i94emxbe3.png" alt=" " width="800" height="726"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;4. Let the Agent Analyze the Annotations&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Once all annotations are complete, click Continue.&lt;/p&gt;

&lt;p&gt;A vision-language model analyzes all of the annotations, and the Agent automatically prepares the parameters and prompt required by the image editing model.&lt;/p&gt;

&lt;p&gt;There is one important implementation detail:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;The Agent sends two images to the image editing model: the original image and a second version containing the visual annotations.&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;This means there is no need to worry about annotation markers covering parts of the original image.&lt;/p&gt;

&lt;p&gt;The model can use the annotated image to understand &lt;strong&gt;where&lt;/strong&gt; the edits should be applied, while still having access to the clean original image for complete, unobstructed visual information.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6dz6mpmu75ifbxhybh3t.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6dz6mpmu75ifbxhybh3t.png" alt=" " width="800" height="461"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;5. Get the Result&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;After these steps, the image editing model generates the final result based on the editing instructions prepared by the Agent.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm9hwa67my3bjpxnb56u3.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm9hwa67my3bjpxnb56u3.png" alt=" " width="800" height="1299"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;This &lt;strong&gt;Annotation + Comment&lt;/strong&gt; approach significantly simplifies the AI image editing workflow.&lt;/p&gt;

&lt;p&gt;Instead of writing complicated prompts, users can directly indicate &lt;strong&gt;where they want to make a change&lt;/strong&gt; and briefly describe &lt;strong&gt;what they want to change&lt;/strong&gt;. The Agent then handles prompt construction and parameter configuration automatically.&lt;/p&gt;

&lt;p&gt;You can deploy PoloX AI locally and test the implementation here:&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/saihhold-zhao/polox_ai" rel="noopener noreferrer"&gt;https://github.com/saihhold-zhao/polox_ai&lt;br&gt;
&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>programming</category>
      <category>tutorial</category>
      <category>opensource</category>
    </item>
  </channel>
</rss>
