HBAI-Ltd/Toonflow-app
HBAI-Ltd/Toonflow-app — Toonflow 是开源一站式 AI 短剧创作工具，将小说、剧本快速转化为动画短剧。集成 AI 编剧、智能分镜、角色与视频生成，跨平台桌面端轻量部署，助力创作者低成本批量产出视觉内容。Toonflow is an open-source A
What is the idea?
An open-source desktop app that carries a novel or script all the way to an animated short drama: AI screenwriting → shot list and storyboard → character and image generation → video assembly, driven by your own model keys. Cross-platform, light deploy, 13k stars in weeks.
Who would use it, and what do they get?
Already a creator product: the buyer is the small 短剧 studio, not a developer.
What is missing today?
It hits a real pain but solves it imperfectly, and it breaks at exactly the step that matters: script → shot list → prompts truncates and silently writes empty panels, so the operator babysits token limits and re-runs. There is also no version for anyone outside 中文 — an English user could not get past the login dialog.
What makes this one different?
Do not build the pipeline. Build the one step that keeps breaking and costs almost nothing to run: a 小说 → 分镜表 converter that never truncates (chunked, resumable, schema-validated) and exports clean, per-shot prompts into 可灵 / 即梦 / Runway — sold per script to the 短剧工作室 who are already paying for generation somewhere else.
What could kill it?
This is the most capital-exposed entry on the board: every second of output costs model money, and 可灵 / 即梦 / 海螺 own both the model and the distribution — ByteDance can fold the whole pipeline into 剪映 and price an indie out overnight. 13k stars sit on nine contributors, so the bus factor is one. The full pipeline is precisely the burn-money-for-speed race the standard excludes; only a narrow, text-only slice of it is open to you.
Who pays, and for what?
短剧工作室 and 小说 IP holders who already pay per generated second — they pay for the step that stops wasting those seconds: a shot list that survives to the last panel.
What is the main risk?
剪映 or 可灵 bundling storyboard generation natively; nine contributors carrying a 13k-star repo.
What did real people actually say?
Verbatim from the public sources linked above, not paraphrased: “每次生成分镜表时，一直被截断，重新生成又被截断，使用的是DeepSeek V4 Pro” — GitHub issue “对话框显示《分镜面板已成功写入12条数据》 但是一个都没看到。导致生成不了” — GitHub issue “docker部署后，浏览器打开页面，非常卡” — GitHub issue
How much work is it?
about two weeks of work
Where does the evidence come from?
https://github.com/HBAI-Ltd/Toonflow-app
← Speechius · Mail Memories – A desktop app to rescue photos from Gmail →