motion_inference / README.md
luizC's picture
Update README.md
caa4e1b verified
---
title: Motion Diffusion Model
emoji: πŸ•Ί
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 6.3.0
app_file: app.py
pinned: false
license: mit
---
# Motion Diffusion Model on Hugging Face
Official implementation of "Human Motion Diffusion Model" by Guy Tevet et al.
## Features
βœ… Text-to-motion generation
βœ… Action-to-motion generation
βœ… Multiple model support
βœ… 3D animation visualization
βœ… Blender export (OBJ, SMPL parameters)
βœ… Fast 50-step models (20x speed)
## Models Available
* **HumanML3D (50-steps)** - Fast text-to-motion
* **HumanML3D (Original)** - Higher quality
* **KIT-ML** - Text-to-motion
* **HumanAct12** - Action-to-motion
* **UESTC** - Action-to-motion
* **Unconstrained** - Free generation
## Usage
1. Select a model type
2. Enter text prompt or action name
3. Adjust parameters
4. Generate animation
5. Export to Blender if needed
## Blender Integration
Exported files can be used with:
* **SMPL add-on** (use JSON parameters)
* **OBJ sequence import** (direct mesh)
* **FBX format** (via conversion)
## Notes
- First run downloads models (~2GB)
- GPU required for optimal performance
- Generation takes 10-30 seconds