For the best experience, it's better to use a desktop computer to view this website.

MemFlow: Flowing Adaptive Memory for Consistent and Efficient Long Video Narratives

Sihui Ji1, Xi Chen1, Shuai Yang3, Xin Tao2, Pengfei Wan2, Hengshuang Zhao1,

1 The University of Hong Kong, 2 Kuaishou Technology, 3 Hong Kong University of Science and Technology (Guangzhou)

Please stay tuned for all video loading...

Hover on the videos to see corresponding text prompts, and drag the progress bar to adjust the playback position.

Contact Us

Feel free to contact Sihui Ji at sihuiji.cs@connect.hku.hk for any question, cooperation, and communication. If you find this work useful, please consider citing:

@misc{ji2025memflow,
                        title={MemFlow: Flowing Adaptive Memory for Consistent and Efficient Long Video Narratives}, 
                        author={Ji, Sihui and Chen, Xi and Yang, Shuai and Tao, Xin and Wan, Pengfei and Zhao, Hengshuang},
                        year={2025},
                        eprint={2512.14699},
                        archivePrefix={arXiv},
                        primaryClass={cs.CV},
                        url={https://arxiv.org/abs/2512.14699}, 
                  }

Thank LongLive, Jenga, UltraPixel, ControlNeXt, and ToonCrafter to provide us the project page's template!