I built a multi-stage pipeline for consistent characters across AI comic panels
Hey Indie Hackers,
I’ve been building Manhwa AI, a tool for generating webtoons from scripts while trying to keep characters consistent across multiple panels.
The workflow is:
script → character setup → storyboard → panel generation → dialogue → export
The hardest part has not been generating a good individual image. It has been maintaining the same face, hairstyle, clothing, body proportions, and visual style across an entire sequence.
To reduce inconsistency, I built the generation process as a multi-stage pipeline instead of sending every panel as an isolated prompt.
The current version lets users:
- Define reusable characters
- Generate and edit storyboards
- Create panels individually
- Add dialogue
- Regenerate failed panels
- Export the completed comic
It’s still an early product, and I’m currently trying to understand which problem matters most to creators:
1. Character consistency
2. Storyboard control
3. Panel editing
4. Faster generation
5. Export and publishing
Here is the product:
I’d appreciate honest feedback, especially from anyone who has experimented with multi-image generation or comics.
