Speculative Decoding for Autoregressive Video Generation

TL;DR

This paper introduces SDVG, a speculative decoding method for autoregressive video diffusion that accelerates inference by using an image-quality router, achieving significant speedups with minimal quality loss.

Contribution

SDVG adapts speculative decoding to autoregressive video generation using an image-quality router, enabling faster inference without architectural changes.

Findings

SDVG retains 98.1% of target quality at 1.59x speedup.

SDVG achieves up to 2.09x speedup with 95.7% quality retention.

SDVG outperforms draft-only generation by over 17%.

Abstract

Autoregressive video diffusion is emerging as a promising paradigm for streaming video synthesis, with step distillation serving as the primary means of accelerating inference. Whether speculative decoding, the dominant acceleration strategy for large language models, can be effectively adapted to autoregressive video generation remains an open question, because video blocks are continuous spatiotemporal tensors with no token-level distribution for exact rejection sampling. We introduce SDVG, which brings speculative decoding to block-based autoregressive video diffusion by replacing token verification with an image-quality router. A 1.3B drafter proposes candidate blocks via four denoising steps; each block is VAE-decoded and scored by ImageReward using worst-frame aggregation--taking the minimum per-frame reward to catch single-frame artifacts that averaging would mask. Blocks scoring above a fixed threshold tau are accepted into the 14B target's KV cache; the rest are regenerated by the target. Two additional design choices prove critical: the first block is always force-rejected to anchor scene composition, and tau serves as a single knob that traces a smooth quality-speed Pareto frontier. On 1003 MovieGenVideoBench prompts (832x480), SDVG retains 98.1% of target-only VisionReward quality (0.0773 vs. 0.0788) at a 1.59x speedup with tau=-0.7, and reaches 2.09x at 95.7% quality retention--while consistently outperforming draft-only generation by over +17%. The framework is training-free, requires no architectural changes, and can be seamlessly integrated into existing autoregressive video generation pipelines.