Human-in-the-loop manga translation studio
A browser workbench that lets fan translators fix text boxes, erase, and redraw manga pages.
- Problem
- Manga fan translators want to translate works they love, but open-source pipelines split text detection, OCR, translation, erasing, and redrawing into separate steps. Furigana at bubble edges get missed, erasing leaves residue, and retouching and lettering remain manual grind. Existing tools are either too academic or not editable.
- Solution
- A semi-automatic browser workbench: upload a manga page, auto-detect text regions that users can add, remove, or adjust, run per-region OCR and translation, erase source text, and redraw with similar fonts. Users edit each block's translation and layout, then export the finished page.
- Users
- Manga scanlation groups, solo fan translators, and creators adding subtitles to foreign comics.
- MVP
- Upload and auto-detect text regions; manually add, remove, and adjust boxes; per-region OCR and translation; erase and simple redraw; edit each block's text and export PNG.
- Revenue
- Free basic tier with watermark, paid watermark removal and batch export; per-page translation credits; team collaboration seat subscription.
- Why now
- Vision models have markedly improved detection and OCR quality, light models can run in the browser, and a solo developer can build an editable semi-automatic workflow that beats fully automatic tools in practice.