Spaces:
Sleeping
Sleeping
File size: 2,483 Bytes
b24a5b7 82a6034 b24a5b7 82a6034 b24a5b7 82a6034 b24a5b7 82a6034 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 |
---
title: ACMDM Motion Generation
emoji: 🎭
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.0.0
app_file: app.py
pinned: false
license: mit
hardware: gpu-t4-small
---
# ACMDM Motion Generation
Generate human motion animations from text descriptions using the ACMDM (Absolute Coordinates Make Motion Generation Easy) model.
## 🎯 Features
- **Text-to-Motion Generation**: Create realistic human motion from natural language descriptions
- **Batch Processing**: Generate multiple motions at once
- **Auto-Length Estimation**: AI automatically determines optimal motion length
- **Flexible Parameters**: Adjust CFG scale, motion length, and more
- **Real-time Preview**: See your generated motions instantly
## 🚀 Usage
1. **Enter a text description** of the motion you want (e.g., "A person is running on a treadmill.")
2. **Adjust parameters** (optional):
- Motion length (40-196 frames)
- CFG scale (controls text alignment)
- Auto-length estimation
3. **Click "Generate Motion"**
4. **View and download** your generated motion video
## 📝 Example Prompts
- "A person is running on a treadmill."
- "Someone is doing jumping jacks."
- "A person walks forward and then turns around."
- "A person is dancing energetically."
## ⚙️ Parameters
- **Motion Length**: Number of frames (40-196). Automatically rounded to multiples of 4.
- **CFG Scale**: Classifier-free guidance scale (1.0-10.0). Higher = more text-aligned, Lower = more diverse.
- **Auto-length**: Let AI estimate the optimal motion length based on your text.
## 🔧 Technical Details
This space uses pre-trained ACMDM models:
- **Autoencoder**: AE_2D_Causal
- **Diffusion Model**: ACMDM_Flow_S_PatchSize22
- **Dataset**: HumanML3D (t2m)
## 📚 Paper
[Absolute Coordinates Make Motion Generation Easy](https://arxiv.org/abs/2505.19377)
## 🤝 Citation
```bibtex
@article{meng2025absolute,
title={Absolute Coordinates Make Motion Generation Easy},
author={Meng, Zichong and Han, Zeyu and Peng, Xiaogang and Xie, Yiming and Jiang, Huaizu},
journal={arXiv preprint arXiv:2505.19377},
year={2025}
}
```
## ⚠️ Notes
- First generation may take 30-60 seconds (model loading)
- Subsequent generations are faster (5-15 seconds)
- GPU recommended for best performance
- Works on CPU but slower
## 🔗 Links
- [GitHub Repository](https://github.com/neu-vi/ACMDM)
- [Project Page](https://neu-vi.github.io/ACMDM/)
- [Paper](https://arxiv.org/abs/2505.19377)
|