GenEraser: Generalizable Video Object Removal via Balanced Text-Mask Guidance and Decoupled Locator-Preserver

Yuqing Chen^{1, 3}, Lin Liu^{2 *}, Haisu Wu⁴, Xiaopeng Zhang², Yaowei Wang^{3, 5}, Yujiu Yang^{1 *}, Qi Tian²

¹Tsinghua University  ²Huawei  ³Pengcheng National Laboratory  ⁴Southeast University  ⁴Harbin Institute of Technology

^*Corresponding Authors

   

GenEraser is a video removal model capable of simultaneously erasing objects and their associated effects. It effectively removes a wide range of physical effects, including smoke, deformations, light, mirrors, shadows, and reflections in open-world scenarios.

Updates

• [2026-5-29] Release arXiv paper.

Showcases

More videos are demonstrated on the project website.

Citation

If you find this work helpful, please help star the repository and consider citing it as follows. It would be greatly appreciated!

@misc{chen2026generasergeneralizablevideoobject,
      title={GenEraser: Generalizable Video Object Removal via Balanced Text-Mask Guidance and Decoupled Locator-Preserver}, 
      author={Yuqing Chen and Lin Liu and Haisu Wu and Xiaopeng Zhang and Yaowei Wang and Yujiu Yang and Qi Tian},
      year={2026},
      eprint={2605.30045},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2605.30045}, 
}

Acknowledgments

We sincerely thank the great work Wan2.2, Wan2.1, VideoX-Fun for their inspiring work and contributions to the video generation community.