EnchantDance: Unveiling the Potential of Music-Driven Dance Movement

EnchantDance: Unveiling the Potential of Music-Driven Dance Movement

Bo Han  Teng Zhang  Zeyu Ling  Yi Ren  Xiang Yin  Feilin Han 
Zhejiang University; Bytedance AI Lab; Beijing Film Academy

🛠️ Environment

Package	Version
Python	3.9.16
PyTorch	1.12.1+cu113
NumPy	1.24.2

Train

🏃 Training

Train VAE

python -m train --cfg configs/config_vae.yaml --cfg_assets configs/assets.yaml --batch_size 64 --nodebug

Train Diffusion

python -m train --cfg configs/config_diffusion.yaml --cfg_assets configs/assets.yaml --batch_size 64 --nodebug

🎯 Inference

🎵 Music to Dance Generation

python demo.py --task=music_dance

🎨 Visualization Examples

• 💃 Motion sequences will be saved as .npz files
• 🎦 Rendered videos to .mp4 files with Blender

🤝 Citation

If you find this repository useful for your work, please consider citing it as follows:

@article{han2023enchantdance,
  title={EnchantDance: Unveiling the Potential of Music-Driven Dance Movement},
  author={Han, Bo and Zhang, Teng and Ling, Zeyu and Ren, Yi and Yin, Xiang and Han, Feilin},
  journal={arXiv preprint arXiv:2312.15946},
  year={2023}
}