Repolode
Submit
blendi-remade

blendi-remade/prompt2rig

Type a text prompt and get a rigged 3D character performing that motion, using fal video models, GVHMR pose recovery and Mixamo retargeting.

View on GitHub
Stars
3
Forks
0
Open issues
0
Last commit
16 hours ago

What is it?

Prompt2Rig turns a text description of a movement into an animated, rigged character in the browser. The prompt is sent to a fast video model on fal (MiniMax H3 Max Turbo or FLUX 3 Max), the human motion in the generated clip is recovered with GVHMR on a local NVIDIA GPU, and the result is retargeted onto a Mixamo skeleton.

A lightweight MediaPipe Pose pass runs in the browser for an immediate live preview, while the higher-quality GVHMR pass produces joint rotations on a fixed-proportion body. Takes can be exported as .glb files with the animation clip, and a headless Blender script renders saved takes to MP4.

The project includes a zero-dependency Node server, a stream mode that chains clips from each clip's last frame, and a FastAPI worker for the HQ pass. The GVHMR and SMPL-X components carry non-commercial terms, so the HQ path is not suitable for commercial use as shipped.

Key features

  • 01Generates a 5-second clip from a text prompt via fal video models.
  • 02Shows a live browser-based pose preview using MediaPipe.
  • 03Recovers motion with GVHMR and retargets it to Mixamo rigs.
  • 04Exports takes as .glb files with the animation clip.
  • 05Stream mode chains clips into one continuous performance.
  • 06Renders saved takes headlessly with Blender.

Use cases

  • Prototyping character animations from text ideas without a mocap setup.
  • Creating preview motion for games or three.js scenes using Mixamo characters.
  • Exporting animated .glb characters for Blender, Unity or three.js.
  • Experimenting with AI video-to-motion pipelines on a local GPU.

Topics

#ai#motion-capture#3d-animation#text-to-video#threejs#pose-estimation#gpu

FAQ

What is prompt2rig?
Type a text prompt and get a rigged 3D character performing that motion, using fal video models, GVHMR pose recovery and Mixamo retargeting. Prompt2Rig turns a text description of a movement into an animated, rigged character in the browser. The prompt is sent to a fast video model on fal (MiniMax H3 Max Turbo or FLUX 3 Max), the human motion in the generated clip is recovered with GVHMR on a local NVIDIA GPU, and the result is retargeted onto a Mixamo skeleton.
Is prompt2rig free to use? What license does it have?
prompt2rig is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is prompt2rig still maintained?
Actively maintained. The last commit to prompt2rig was on Oct 19, 2026.
What can I use prompt2rig for?
Prototyping character animations from text ideas without a mocap setup. Creating preview motion for games or three.js scenes using Mixamo characters. Exporting animated .glb characters for Blender, Unity or three.js. Experimenting with AI video-to-motion pipelines on a local GPU.
Where can I get prompt2rig?
The source code and installation instructions are on GitHub: https://github.com/blendi-remade/prompt2rig

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects