Video diffusion accelerated with historical trajectory guidance
TGuide: Faster Autoregressive Video Diffusion via Historical Trajectory Guidance
Computer Vision and Pattern Recognition
Summary
Generating videos by gradually refining noisy images is slow, especially toward the last few steps. The authors found that features from previous video parts can help predict changes more quickly and accurately. They designed a method called TGuide that uses information from both the current and past video chunks to speed up this process. This approach keeps video quality high while almost doubling the speed of video generation.
What this means in practice
- •For video generation engineers: Speed up autoregressive video diffusion for faster video synthesis without quality loss.
- •For game developers: Generate longer realistic video sequences efficiently for gameplay or simulation using autoregressive diffusion models.
Authors
Zhuo Chen, Bin Zou, Fanyue Wei, Lixin Duan, Angela Yao, Wen Li
Abstract
Can autoregressive video diffusion be accelerated further when only four denoising steps remain? Existing feature caching approaches struggle in this regime, as large feature changes between denoising timesteps undermine both direct reuse and lightweight predictors. We observe that features across video chunks remain strongly correlated at matching denoising timesteps, offering a complementary source of guidance. Building on this insight, we introduce TGuide, a framework that accelerates autoregressive video diffusion through a lightweight Anchor-Transport-Correct (ATC) predictor guided by historical sampling trajectories. Our ATC combines two complementary references: the current chunk's preceding-step features preserve its spatial structure, while the previous chunk's same-timestep features provide information at the target denoising stage. By integrating these references through spatial retrieval and adaptive correction, ATC improves velocity prediction accuracy despite the large gaps between denoising steps. To mitigate error accumulation across both denoising steps and video chunks, we train the predictor on its own autoregressive rollouts, allowing it to adapt to states shaped by its own earlier predictions. An optional confidence-aware scheduler further adjusts the quality-efficiency trade-off. Experiments on Self Forcing, Causal Forcing, and HY-WorldPlay show that TGuide preserves generation quality while achieving denoising speedups of up to $1.89\times$ over the original 4-step generators. On Self Forcing, TGuide generalizes from 81-frame training videos to 333-frame generation, approximately four times the training length, while exceeding the original model's VBench score.