Dev open-sources ComfyUI movie studio that turns one prompt into a fully editable film
Goat_bless · reddit · 2026-10-21
A Reddit user has open-sourced their ComfyUI film-creation and retouching studio (GitHub: SpendinFR/H3STUDIOFILM). One prompt (e.g. "aliens invade the world") flows through a director-style pipeline: content director creates acts/scenes → stage director expands scenes → plan director breaks shots → clip director generates each clip, with stepwise refinement and assembly into a final film.
- Supports visual references, including generating them live via Krea.
- Everything is controllable manually or via natural language ("change this part, put this instead"); retouch mode covers delete/replace characters, change camera angle or decor, area retouch, precise retakes, add elements.
- The author notes it's imperfect, bugs remain, and ComfyUI configs are hard to replicate across machines — follow the README and enlist an AI agent to adapt it.
More from Multimodal
- Ghostly AI breakup comedy short showcases steady character consistency — SimplySarah_06 · 2026-10-22
- One Selfie Video, Many Worlds: Creator Transfers Real Motion Into AI Motion Graphics Scenes — aziz4ai · 2026-10-22
- Mollick: Opus 5.5 turned his otter-on-a-plane test into a video and picked his ElevenLabs voice clone on its own — emollick · 2026-10-22
- Maker blends ComfyUI, Seedance API, MiniMax H3 and Suno in a 1:40 AI music video — alecubudulecu · 2026-10-22
- Mystery video model Topview T1 claims Seedance 2.5-level 1080p at 62% lower cost; early tests impress — iamfakhrealam · 2026-10-22
- Gradium open-sources Phonon: 64MB on-device TTS with ~7ms first audio, 37 voices — lmoroney · 2026-10-22