Anchor Forcing: Anchor Memory and Tri-Region RoPE for Interactive Streaming Video Diffusion
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911514607222784 |
|---|---|
| author | Yang, Yang Zhang, Tianyi Huang, Wei Chen, Jinwei Wu, Boxi He, Xiaofei Cai, Deng Li, Bo Jiang, Peng-Tao |
| author_facet | Yang, Yang Zhang, Tianyi Huang, Wei Chen, Jinwei Wu, Boxi He, Xiaofei Cai, Deng Li, Bo Jiang, Peng-Tao |
| contents | Interactive long video generation requires prompt switching to introduce new subjects or events, while maintaining perceptual fidelity and coherent motion over extended horizons. Recent distilled streaming video diffusion models reuse a rolling KV cache for long-range generation, enabling prompt-switch interaction through re-cache at each switch. However, existing streaming methods still exhibit progressive quality degradation and weakened motion dynamics. We identify two failure modes specific to interactive streaming generation: (i) at each prompt switch, current cache maintenance cannot simultaneously retain KV-based semantic context and recent latent cues, resulting in weak boundary conditioning and reduced perceptual quality; and (ii) during distillation, unbounded time indexing induces a positional distribution shift from the pretrained backbone's bounded RoPE regime, weakening pretrained motion priors and long-horizon motion retention. To address these issues, we propose \textbf{Anchor Forcing}, a cache-centric framework with two designs. First, an anchor-guided re-cache mechanism stores KV states in anchor caches and warm-starts re-cache from these anchors at each prompt switch, reducing post-switch evidence loss and stabilizing perceptual quality. Second, a tri-region RoPE with region-specific reference origins, together with RoPE re-alignment distillation, reconciles unbounded streaming indices with the pretrained RoPE regime to better retain motion priors. Experiments on long videos show that our method improves perceptual quality and motion metrics over prior streaming baselines in interactive settings. Project page: https://github.com/vivoCameraResearch/Anchor-Forcing |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_13405 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Anchor Forcing: Anchor Memory and Tri-Region RoPE for Interactive Streaming Video Diffusion Yang, Yang Zhang, Tianyi Huang, Wei Chen, Jinwei Wu, Boxi He, Xiaofei Cai, Deng Li, Bo Jiang, Peng-Tao Computer Vision and Pattern Recognition Image and Video Processing Interactive long video generation requires prompt switching to introduce new subjects or events, while maintaining perceptual fidelity and coherent motion over extended horizons. Recent distilled streaming video diffusion models reuse a rolling KV cache for long-range generation, enabling prompt-switch interaction through re-cache at each switch. However, existing streaming methods still exhibit progressive quality degradation and weakened motion dynamics. We identify two failure modes specific to interactive streaming generation: (i) at each prompt switch, current cache maintenance cannot simultaneously retain KV-based semantic context and recent latent cues, resulting in weak boundary conditioning and reduced perceptual quality; and (ii) during distillation, unbounded time indexing induces a positional distribution shift from the pretrained backbone's bounded RoPE regime, weakening pretrained motion priors and long-horizon motion retention. To address these issues, we propose \textbf{Anchor Forcing}, a cache-centric framework with two designs. First, an anchor-guided re-cache mechanism stores KV states in anchor caches and warm-starts re-cache from these anchors at each prompt switch, reducing post-switch evidence loss and stabilizing perceptual quality. Second, a tri-region RoPE with region-specific reference origins, together with RoPE re-alignment distillation, reconciles unbounded streaming indices with the pretrained RoPE regime to better retain motion priors. Experiments on long videos show that our method improves perceptual quality and motion metrics over prior streaming baselines in interactive settings. Project page: https://github.com/vivoCameraResearch/Anchor-Forcing |
| title | Anchor Forcing: Anchor Memory and Tri-Region RoPE for Interactive Streaming Video Diffusion |
| topic | Computer Vision and Pattern Recognition Image and Video Processing |
| url | https://arxiv.org/abs/2603.13405 |