Novel to Continuous Comic

Problem

Sending a novel, web story, or chapter directly to an image model usually produces pages with drifting character designs, weak storyboard pacing, and unrelated illustrations. novel-to-comic treats comic adaptation as a staged pipeline: analyze the source, define characters, write a storyboard and per-page prompts, then generate images and merge them into a PDF.

How It Works and Limits

Its main value is continuity control rather than raw image generation. It recommends art, tone, layout, or presets from content signals, such as ohmsha for tutorials, wuxia for martial arts, and four-panel for short fables. Projects are written to comic/{topic-slug}/, preserving analysis.md, storyboard.md, characters/characters.md, prompts/*.md, and generated pages for review and regeneration.
Key steps include:
- Config check: read EXTEND.md first; missing setup blocks later workflow.
- Content confirmation: decide focus, audience, --art, --tone, and --layout.
- Partial generation: use --storyboard-only, --prompts-only, --images-only, or --regenerate N to work on one stage.
- Continuity anchors: prefer the latest approved page as the main reference, then the character sheet, so a prettier page that breaks identity is treated as a failed result.
It depends on an installed image generation skill, and --ref failures may require compressed references or a prompt-only fallback. It fits multi-page narrative, chapter adaptation, and stable-character work, not single-image requests where storyboard and prompt review are unnecessary.