Motion-Aware Caching for Efficient Autoregressive Video Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Jing, Ma, Yuexiao, Zheng, Xuzhe, Wang, Xing, Liu, Shiwei, Yan, Chenqian, Zheng, Xiawu, Ji, Rongrong, Chao, Fei, Liu, Songwei
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911685275549696
author Xu, Jing
Ma, Yuexiao
Zheng, Xuzhe
Wang, Xing
Liu, Shiwei
Yan, Chenqian
Zheng, Xiawu
Ji, Rongrong
Chao, Fei
Liu, Songwei
author_facet Xu, Jing
Ma, Yuexiao
Zheng, Xuzhe
Wang, Xing
Liu, Shiwei
Yan, Chenqian
Zheng, Xiawu
Ji, Rongrong
Chao, Fei
Liu, Songwei
contents Autoregressive video generation paradigms offer theoretical promise for long video synthesis, yet their practical deployment is hindered by the computational burden of sequential iterative denoising. While cache reuse strategies can accelerate generation by skipping redundant denoising steps, existing methods rely on coarse-grained chunk-level skipping that fails to capture fine-grained pixel dynamics. This oversight is critical: pixels with high motion require more denoising steps to prevent error accumulation, while static pixels tolerate aggressive skipping. We formalize this insight theoretically by linking cache errors to residual instability, and propose MotionCache, a motion-aware cache framework that exploits inter-frame differences as a lightweight proxy for pixel-level motion characteristics. MotionCache employs a coarse-to-fine strategy: an initial warm-up phase establishes semantic coherence, followed by motion-weighted cache reuse that dynamically adjusts update frequencies per token. Extensive experiments on state-of-the-art models like SkyReels-V2 and MAGI-1 demonstrate that MotionCache achieves significant speedups of $\textbf{6.28}\times$ and $\textbf{1.64}\times$ respectively, while effectively preserving generation quality (VBench: $1\%\downarrow$ and $0.01\%\downarrow$ respectively). The code is available at https://github.com/ywlq/MotionCache.
format Preprint
id arxiv_https___arxiv_org_abs_2605_01725
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Motion-Aware Caching for Efficient Autoregressive Video Generation
Xu, Jing
Ma, Yuexiao
Zheng, Xuzhe
Wang, Xing
Liu, Shiwei
Yan, Chenqian
Zheng, Xiawu
Ji, Rongrong
Chao, Fei
Liu, Songwei
Computer Vision and Pattern Recognition
Artificial Intelligence
Autoregressive video generation paradigms offer theoretical promise for long video synthesis, yet their practical deployment is hindered by the computational burden of sequential iterative denoising. While cache reuse strategies can accelerate generation by skipping redundant denoising steps, existing methods rely on coarse-grained chunk-level skipping that fails to capture fine-grained pixel dynamics. This oversight is critical: pixels with high motion require more denoising steps to prevent error accumulation, while static pixels tolerate aggressive skipping. We formalize this insight theoretically by linking cache errors to residual instability, and propose MotionCache, a motion-aware cache framework that exploits inter-frame differences as a lightweight proxy for pixel-level motion characteristics. MotionCache employs a coarse-to-fine strategy: an initial warm-up phase establishes semantic coherence, followed by motion-weighted cache reuse that dynamically adjusts update frequencies per token. Extensive experiments on state-of-the-art models like SkyReels-V2 and MAGI-1 demonstrate that MotionCache achieves significant speedups of $\textbf{6.28}\times$ and $\textbf{1.64}\times$ respectively, while effectively preserving generation quality (VBench: $1\%\downarrow$ and $0.01\%\downarrow$ respectively). The code is available at https://github.com/ywlq/MotionCache.
title Motion-Aware Caching for Efficient Autoregressive Video Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.01725