A few months ago I started experimenting with a strange idea:
What if AI could generate an actual readable webtoon episode instead of just random anime images?
Not just “AI art.”
An actual vertical story people would want to scroll through.
At first, I assumed this would mostly be an image generation problem.
I thought the hard parts would be:
- generating attractive anime-style characters
- making panels visually polished
- creating cinematic backgrounds
- improving image consistency
But after building prototype after prototype, I realized something unexpected:
The hardest part of AI webtoons has almost nothing to do with image quality.
It’s storytelling.
More specifically:
- pacing
- sequencing
- emotional continuity
- panel transitions
- dialogue flow
- scroll rhythm
Modern AI models are already very good at generating single images.
You can prompt:
“A fantasy heroine standing in the rain at night”
…and instantly get something visually impressive.
But webtoons are fundamentally different from standalone AI images.
A webtoon is not one image.
It’s rhythm.
That was the biggest thing I underestimated before building BuildToon.
Vertical storytelling changes everything
Traditional comics and webtoons are consumed completely differently.
In a comic page, the reader sees multiple panels simultaneously.
In a webtoon, the pacing is controlled by scrolling.
That changes:
- suspense
- reveals
- pauses
- emotional timing
- conversations
- scene transitions
A technically beautiful panel can completely fail if the transition to the next panel feels awkward.
And surprisingly, users noticed this immediately.
I started seeing that people cared much more about:
- emotional pacing
- readable dialogue
- story flow
- continuity
…than “perfect” image quality.
That honestly surprised me.
Short prompts worked better than giant prompts
Another unexpected discovery:
Huge lore dumps often produced worse results.
When users pasted giant 5,000-word fantasy documents, episodes became less coherent.
But simpler prompts like:
“Two former childhood friends reunite during a rainy midnight train ride after years apart.”
…often created much stronger webtoon episodes.
The AI handled emotional direction much better when the core concept itself was emotionally focused.
That completely changed how I started thinking about AI storytelling and prompt design.
Character consistency is still incredibly hard
Generating one anime image is easy now.
Generating:
- 8 connected webtoon panels
- with the same characters
- stable clothing
- consistent faces
- readable composition
- believable emotional progression
- coherent scene flow
…is a completely different problem.
Sometimes the AI produced:
- beautiful standalone panels
- but terrible storytelling
Other times:
- mediocre art
- but surprisingly engaging pacing
And users consistently preferred the second one.
That was probably the most important lesson so far:
People forgive imperfect visuals much faster than they forgive broken storytelling.
The analytics changed how I think about the product
I originally assumed users would:
- generate one episode
- export it
- leave forever
But one of the most clicked actions unexpectedly became:
“Continue story.”
That single interaction completely changed how I think about AI-native storytelling.
It made me realize users are not just looking for AI-generated images.
They want:
- recurring characters
- serialized stories
- evolving relationships
- emotional continuity
- persistent worlds
- story memory
In other words:
They want the same things readers already love about traditional webtoons.
The AI part alone is not enough.
What people actually care about is whether the story feels alive.
SEO started working earlier than expected
Another weird moment:
Google started indexing pages around topics like:
- AI webtoon generator
- how to make a webtoon
- webtoon creator
- AI comic creator
- webtoon panel generator
…much earlier than I expected.
That was surprising because internally the project still felt extremely early.
Most traffic currently comes from countries where webtoons already have massive traction:
- South Korea
- Indonesia
- Thailand
- Singapore
Which honestly makes perfect sense in hindsight.
I think AI storytelling is still massively unsolved
AI image generation is already becoming commoditized.
Generating beautiful images is no longer enough.
But generating stories people emotionally care about?
That still feels incredibly unsolved.
Right now I’m experimenting with:
- story continuation
- recurring characters
- lightweight editing
- dialogue rewriting
- panel regeneration
- mobile reading flow
- TikTok-ready exports
- AI-assisted storytelling workflows
The overlap between:
- AI storytelling
- vertical media
- webtoons
- short-form content
- creator tools
…feels much bigger than I initially realized.
Especially now that platforms like TikTok, Reels, and Shorts have trained people to consume narrative content vertically.
In some strange way, webtoons were already optimized for the modern internet long before AI entered the picture.
I’m still very early in this journey and still figuring out what this product ultimately becomes.
But building an AI webtoon generator has easily been one of the most fascinating technical and creative problems I’ve worked on.
Curious to see where AI-native storytelling goes over the next few years.
Would genuinely love feedback from:
- webtoon readers
- comic artists
- indie hackers
- AI builders
- people experimenting with storytelling tools
Still experimenting with all of this:
https://www.buildtoon.com
Top comments (0)