# I built a free, open-source film studio on top of ComfyUI + MiniMax H3: idea to finished short film, on your own PC — retained record

> Retained reddit record from reddit:comfyui.

Canonical: https://highsignal.app/data/records/3dc2146e37cf705b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:comfyui)  published 2026-11-03 10:10 ingested 2026-11-04 02:41

[original source ↗](https://www.reddit.com/r/comfyui/comments/1xmp6dh/i_built_a_free_opensource_film_studio_on_top_of/)

## retained text



```
This is IV Studio, my entry for Comfy's Open Source Higgsfield challenge. It's a Windows desktop app (MIT) that drives your own ComfyUI over its API, so you never touch a node graph unless you want to.

What it does:

\- \*\*Write it for me:\*\* a local model (Qwen3-VL via Ollama or your AI Agent through MCP) looks at your reference pictures and writes the H3 prompt. A checker catches the mistakes H3 punishes (frame-grid lengths, speaker tags, unattached references) and fixes most in one click.

\- \*\*Hunt, then finish:\*\* quick previews with live step pictures; takes are ranked by what OpenAI Whisper hears, likeness and sharpness. A review names each problem with its timestamp. The pick finishes from the same latent with DLSS 5 or SeedVR2.

\- \*\*Projects:\*\* a six-step wizard from idea to short film (script, cast, storyboard sketches you can play as an animatic). Every shot is filmed from its sketch, so faces and rooms stay consistent.

\- \*\*Staging board:\*\* place people and the camera on a top-down floor plan; it becomes exact screen directions, or a grey previs clip H3 follows for arcing moves.

\- \*\*An editor that makes footage:\*\* right-click to extend a shot, fill a gap, retake a line, cut someone out (SAM 3.1) or reframe to vertical.

\- \*\*Also in there:\*\* character LoRA training, Replace (Genjutsu Level), music (MiniMax Music 3 / ACE-Step), local voices (VoxCPM2), Civitai LoRAs, and a 142-tool MCP server so Claude Code or any agent can direct alongside you.

\- \*\*No strong GPU?\*\* It can render on your own Runpod or Modal account and shuts the GPU off when idle. The Runpod path is brand new, so feedback is welcome.

Tested on a 4090 with 64 GB RAM; 16 GB cards work with smaller previews. It downloads each model from its host and checks its SHA-256.

Download: [https://github.com/hollisorg/iv-studio/releases/latest](https://github.com/hollisorg/iv-studio/releases/latest)

Code: [https://github.com/hollisorg/iv-studio](https://github.com/hollisorg/iv-studio)

Happy to answer questions about how it builds and drives the graphs.

Edit: Now has Linux support!
```



document fetched 2026-11-04 00:41 · canonical https://reddit.com/r/comfyui/comments/1xmp6dh/i_built_a_free_opensource_film_studio_on_top_of
