UniTransfer: Video Concept Transfer via Progressive Spatial and Timestep Decomposition (NeurIPS 2025)

• arXiv
• Paper
• Huggingface
• Project Page

OverView





Project Updates

• 🔥🔥 We finished foreground transfer
• 🔥🔥 We finished the background transfer to guide the video generation
• 🔥🔥 We finished human animation
• 🔥🔥 We finshed animal transfering from a reference image
• ......... And More

Quick Start

Diffusers

Please make sure your Python version is between 3.10 and 3.12, inclusive of both 3.10 and 3.12.

conda env create -f environment.yml
python infer.py

Gallery



Acknowledgement

• CogVideoX
• MagicMotion
• Tora

Thanks CogVideoX ,MagicMotion, Tora for their great contribution to the AIGC

Citation

🌟 If you find our work helpful, please leave us a star and cite our paper. Our Code is mainly based on CogVideoX，Please Cite Them Too.

@misc{lei2025unitransfervideoconcepttransfer,
      title={UniTransfer: Video Concept Transfer via Progressive Spatial and Timestep Decomposition}, 
      author={Guojun Lei and Rong Zhang and Chi Wang and Tianhang Liu and Hong Li and Zhiyuan Ma and Weiwei Xu},
      year={2025},
      eprint={2509.21086},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2509.21086}, 
}

Model-License

The code in this repository is released under the Apache 2.0 License.