Adaptive Begin-of-Video Tokens for Autoregressive Video Diffusion Models

TL;DR

This paper introduces adaptive begin-of-video tokens for autoregressive video diffusion models, enhancing long video generation by improving global consistency and local dynamics through learnable embeddings and a refinement strategy.

Contribution

The paper proposes a novel adaptive BOV token mechanism and a refinement strategy for stream denoising, advancing long video generation quality and consistency in diffusion models.

Findings

Achieves better global consistency in long video synthesis.

Improves local motion dynamics and image quality.

Demonstrates superior performance on multiple metrics.

Abstract

Recent advancements in diffusion-based video generation have produced impressive and high-fidelity short videos. To extend these successes to generate coherent long videos, most video diffusion models (VDMs) generate videos in an autoregressive manner, i.e., generating subsequent frames conditioned on previous ones. There are generally two primary paradigms: chunk-based extension and stream denoising. The former directly concatenates previous clean frames as conditioning, suffering from denoising latency and error accumulation. The latter maintains the denoising sequence with monotonically increasing noise levels. In each denoising iteration, one clean frame is produced while a new pure noise is simultaneously appended, enabling live-stream sampling. However, it struggles with fragile consistency and poor motion dynamics. In this paper, we propose Adaptive Begin-of-Video Tokens (ada-BOV) for autoregressive VDMs. The BOV tokens are special learnable embeddings on VDMs. They adaptively absorb denoised preceding frames via an adaptive-layer-norm-like modulation. This design preserves the global consistency while allowing for flexible conditioning in dynamic scenarios. To ensure the quality of local dynamics essential in modulating BOV tokens, we further propose a refinement strategy for stream denoising. It decouples the sampling trajectory length from the attention window size constraint, leading to improved local guidance and overall imaging quality. We also propose a disturbance-augmented training noise schedule, which balances the convergence speed with model robustness for the stream denoising. Extensive experiments demonstrate that our method achieves compelling qualitative and quantitative results across multiple metrics.