Not All Frames Deserve Full Computation: Accelerating Autoregressive Video Generation via Selective Computation and Predictive Extrapolation

TL;DR

SCOPE is a training-free framework that accelerates autoregressive video diffusion by selectively choosing between cache, prediction, and recomputation, using noise-level extrapolation and active frame restriction.

Contribution

It introduces a tri-modal scheduler and selective computation techniques that improve efficiency without training, addressing AR-specific inefficiencies.

Findings

Achieves up to 4.73x speedup on MAGI-1 and SkyReels-V2 datasets.

Maintains comparable video quality to original methods.

Outperforms all existing training-free acceleration baselines.

Abstract

Autoregressive (AR) video diffusion models enable long-form video generation but remain expensive due to repeated multi-step denoising. Existing training-free acceleration methods rely on binary cache-or-recompute decisions, overlooking intermediate cases where direct reuse is too coarse yet full recomputation is unnecessary. Moreover, asynchronous AR schedules assign different noise levels to co-generated frames, yet existing methods process the entire valid interval uniformly. To address these AR-specific inefficiencies, we present SCOPE, a training-free framework for efficient AR video diffusion. SCOPE introduces a tri-modal scheduler over cache, predict, and recompute, where prediction via noise-level Taylor extrapolation fills the gap between reuse and recomputation with explicit stability controls backed by error propagation analysis. It further introduces selective computation that restricts execution to the active frame interval. On MAGI-1 and SkyReels-V2, SCOPE achieves up to 4.73x speedup while maintaining quality comparable to the original output, outperforming all training-free baselines.