[NeurIPS 2025] OmniVCus: Feedforward Subject-driven Video Customization with Multimodal Control Conditions

Model Description

These three models support multi-modal control video customization tasks, including reference-to-video, reference-mask-to-video, reference-depth-to-video, and reference-instruction-to-video generation. Our models are based on Wan2.1-1.3B, Wan2.1-14B, Wan2.2-14B, and VACE. Here are some comparisons with the state-of-the-art method VACE on video customization:

· (a) 2.1-1.3B model

· (b) 2.1-14B model

· (c) 2.2-14B model

Github Code Link

Please refer to our GitHub repo for more detailed instructions on using our code and models.

https://github.com/caiyuanhao1998/Open-OmniVCus

Training Data Link

Our models are trained on our curated dataset:

https://huggingface.co/datasets/CaiYuanhao/OmniVCus-Train

Testing Data Link

We provide 648 data samples to test our models

https://huggingface.co/datasets/CaiYuanhao/OmniVCus-Test

Project Page Link

For more video customization results, please refer to our project page:

https://caiyuanhao1998.github.io/project/OmniVCus/

Arxiv Paper Link

For more technical details, please refer to our NeurIPS 2025 paper:

https://arxiv.org/abs/2506.23361

Citation

If you find our code, data, and models useful, please consider citing our paper:

@inproceedings{omnivcus,
  title={OmniVCus: Feedforward Subject-driven Video Customization with Multimodal Control Conditions},
  author={Yuanhao Cai and He Zhang and Xi Chen and Jinbo Xing and Kai Zhang and Yiwei Hu and Yuqian Zhou and Zhifei Zhang and Soo Ye Kim and Tianyu Wang and Yulun Zhang and Xiaokang Yang and Zhe Lin and Alan Yuille},
  booktitle={NeurIPS},
  year={2025}
}