UltraViCo: Breaking Extrapolation Limits in Video Diffusion Transformers

TL;DR

UltraViCo is a training-free method that enhances video diffusion transformers' ability to generalize beyond training length by suppressing attention dispersion, significantly improving extrapolation performance and quality.

Contribution

We introduce UltraViCo, a novel plug-and-play approach that addresses attention dispersion to improve video length extrapolation in diffusion transformers without additional training.

Findings

Outperforms baselines across models and ratios

Increases extrapolation limit from 2x to 4x

Improves quality metrics by over 40% at 4x extrapolation

Abstract

Despite advances, video diffusion transformers still struggle to generalize beyond their training length, a challenge we term video length extrapolation. We identify two failure modes: model-specific periodic content repetition and a universal quality degradation. Prior works attempt to solve repetition via positional encodings, overlooking quality degradation and achieving only limited extrapolation. In this paper, we revisit this challenge from a more fundamental view: attention maps, which directly govern how context influences outputs. We identify that both failure modes arise from a unified cause: attention dispersion, where tokens beyond the training window dilute learned attention patterns. This leads to quality degradation and repetition emerges as a special case when this dispersion becomes structured into periodic attention patterns, induced by harmonic properties of positional encodings. Building on this insight, we propose UltraViCo, a training-free, plug-and-play method that suppresses attention for tokens beyond the training window via a constant decay factor. By jointly addressing both failure modes, we outperform a broad set of baselines largely across models and extrapolation ratios, pushing the extrapolation limit from 2x to 4x. Remarkably, it improves Dynamic Degree and Imaging Quality by 233% and 40.5% over the previous best method at 4x extrapolation. Furthermore, our method generalizes seamlessly to downstream tasks such as controllable video synthesis and editing.