Motion-Aware Caching for Efficient Autoregressive Video Generation

TL;DR

MotionCache introduces a motion-aware caching strategy for autoregressive video generation, significantly speeding up process while maintaining quality by dynamically adjusting denoising steps based on pixel motion.

Contribution

It formalizes the link between cache errors and residual instability and proposes a lightweight, motion-aware cache framework with a coarse-to-fine strategy for improved efficiency.

Findings

Achieves 6.28x speedup on SkyReels-V2

Achieves 1.64x speedup on MAGI-1

Preserves generation quality with minimal degradation

Abstract

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.