StreamErase: Towards Ultra-Fast Streaming Video Erasing

Zhuobai Dong 1, Xiaobing Tu 2, Xiantao Zhang 2, Jinkui Ren 2, Yinggui Wang 2, Yue Ma 3, Linfeng Zhang 4
1WuHan University,    2Alibaba Group,    3The Hong Kong University of Science and Technology,
4Shanghai Jiao Tong University
 
Paper   Code   Hugging Face Model
Click to zoom in

Prior video diffusion erasing models typically operate offline to generate fixed-length sequences in parallel (top left). In contrast, our StreamErase enables streaming long-video generation (bottom left), supporting real-time erasing at 26.06 FPS with 0.87s latency (right).

Qualitative Comparison

Removed results comparison with background video

Short Videos

Original Video
StreamErase
Original Video
StreamErase
Loading short videos...

Long Videos

Original Video
StreamErase
Loading long videos...
Loading video comparisons...
Videos are synchronized within each original/generated pair. Play, pause, or seek either video to compare them.

BibTeX 

@misc{dong2026streameraseultrafaststreamingvideo,
      title={StreamErase: Towards Ultra-Fast Streaming Video Erasing}, 
      author={Zhuobai Dong and Xiaobing Tu and Xiantao Zhang and Jinkui Ren and Yinggui Wang and Yue Ma and Linfeng Zhang},
      year={2026},
      eprint={2610.20137},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2610.20137}, 
}