README.md

metadata

license: mit
language:
  - en
base_model:
  - Wan-AI/Wan2.1-T2V-1.3B
pipeline_tag: text-to-video
tags:
  - Real-Time
  - Long-Video
  - Video-Diffusion-Model
  - Autoregressive

Rolling Forcing

Autoregressive Long Video Diffusion in Real Time

Kunhao Liu¹ · Wenbo Hu² · Jiale Xu² · Ying Shan² · Shijian Lu^1
1Nanyang Technological University ²ARC Lab, Tencent PCG

💡 TL;DR: REAL-TIME streaming generation of MULTI-MINUTE videos

• 🚀 Real-Time at 16 FPS: Stream high-quality video directly from text on a single GPU.
• 🎬 Minute-Long Videos: Generate coherent, multi-minute sequences with dramatically reduced drift.
• ⚙️ Rolling-Window Strategy: Denoise frames together in a rolling window for mutual refinement, breaking the chain of error accumulation.
• 🧠 Long-Term Memory: The novel Attention Sink anchors your video, preserving global context over thousands of frames.
• 🥇 State-of-the-Art Performance: Outperforms all comparable open-source models in quality and consistency.

📚 Citation

If you find this codebase useful for your research, please cite our paper and consider giving the repo a ⭐️ on GitHub: https://github.com/TencentARC/RollingForcing

@article{liu2025rolling,
  title={Rolling Forcing: Autoregressive Long Video Diffusion in Real Time},
  author={Liu, Kunhao and Hu, Wenbo and Xu, Jiale and Shan, Ying and Lu, Shijian},
  journal={arXiv preprint arXiv:2509.25161},
  year={2025}
}