Hi HN, I'm Sanjoy. I've always loved comic books and stories but can't draw. So I built ComicInk: describe a story, the AI generates a complete comic book — consistent characters across panels, dialogue, cover art, the whole thing.
No-signup demo at comicink.ai/quick — type a prompt, watch Gemini write a dramatic teaser of your story in real-time while the cover + page 1 render (~2 min). Free signup unlocks the full 4-page comic.
The hard problem was character consistency. Most AI image tools treat each generation independently, so your hero's face morphs between panels. ComicInk stores per-character reference images + structured attributes (age group, build, hair, distinguishing features) and injects them into every panel prompt. Not perfect, but substantially better than raw text-to-image.
Translation worked out better than I expected: a click renders the comic in 10+ languages while the panel art stays identical — only the dialogue and captions change. The public reader auto-detects the visitor's browser language and serves the matching edition if one exists.
Getting some interest from serious book creators as well so I introduced the "Book to comic" feature. Simply upload the PDF of a book of you have written. AI reads the book, analyzes it for important plot elements and characters. You can decide to convert the book into 1 issue or split it up across multiple issues, choose the art style you want and you have a comic version of your book.
Stack: Next.js 16 / Supabase / Gemini (text + images) / fal.ai (backup image provider) / Stripe / Inngest. iOS app on the App Store.
Happy to answer anything. Especially curious if anyone's solved character consistency better than reference-image injection — it holds up reasonably well with the Gemini image models I use, but drifts on other providers.
Comments URL: https://news.ycombinator.com/item?id=47964060
Points: 1
# Comments: 2