femnn/ai-3d-director-desk — explained in plain English
Analysis updated 2026-05-18
Pre-visualize a film or video scene in 3D before shooting it for real
Turn several phones into independent virtual cameras for a multi-angle shoot
Pose and animate characters using preset motions, video extraction, or text prompts
Record short camera movement clips and export them directly as MP4
| femnn/ai-3d-director-desk | aerdelan/housand-domaintoolmatrix | affaan-m/opencode | |
|---|---|---|---|
| Stars | 22 | 22 | 22 |
| Language | TypeScript | TypeScript | TypeScript |
| Last pushed | — | — | 2026-02-09 |
| Maintenance | — | — | Maintained |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | vibe coder | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Runs locally with npm install and npm run dev, or download a packaged desktop build from Releases.
AI 3D Director Desk is a local-first tool for planning and pre-visualizing film and video scenes in 3D. Instead of building everything on a single computer screen, it turns spare phones into virtual cameras that people can walk around a scene with, while a computer runs the main 3D set. A user builds a 3D scene on a computer, adding characters, props, and camera positions, either by hand or by importing a structured set of AI-generated scene commands. An AI agent can add characters, props, and cameras using a fixed list of allowed actions, so it never runs arbitrary code, only pre-approved commands described in JSON. Once the scene exists, other people join from their phones by scanning a QR code shown on the director desk. Each phone becomes its own independent camera, controlled with an on-screen joystick to move and a drag gesture to aim, plus controls for height and zoom. The desktop and the connected phones all show the same live camera view at once, so a director sitting at the computer can watch every phone's shot in real time. The tool can record 5, 10, or 15 second clips of a camera's movement and export the result straight to MP4, keeping both the recorded video and the camera's path so it can be replayed or deleted later. Characters can be posed and animated in several ways: looping preset motions, motion driven by the camera's own movement, motion extracted from an uploaded video or image, direct editing of the character's skeleton, and a text-to-motion feature called AnimoFlow that turns written prompts into animation. Everything can run on a local network without sending data anywhere, and project files are only uploaded if the user chooses to export or share them. The project can be run from source with Node.js and npm, or downloaded as a packaged desktop app for macOS on Apple Silicon or Windows. It is built with React, TypeScript, Three.js, and Electron, and is an independently developed continuation of an earlier open source project by another author. It is released under the MIT license, meaning it can be used freely, including for commercial purposes.
A local-first 3D scene planning tool that turns phones into virtual cameras and records the shots straight to MP4.
Mainly TypeScript. The stack also includes TypeScript, React, Three.js.
MIT license: use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.