TS-Attn: Temporal-wise Separable Attention for Multi-Event Video Generation

TL;DR

This paper introduces TS-Attn, a novel attention mechanism that improves multi-event video generation by enhancing temporal coherence and content alignment in pre-trained models.

Contribution

The paper proposes a training-free, plug-and-play attention method, TS-Attn, that enhances temporal awareness and coherence in multi-event video synthesis.

Findings

Boosts StoryEval-Bench scores by over 33% on Wan2.1-T2V-14B.

Increases inference time by only 2%.

Supports multi-event image-to-video generation across models.

Abstract

Generating high-quality videos from complex temporal descriptions that contain multiple sequential actions is a key unsolved problem. Existing methods are constrained by an inherent trade-off: using multiple short prompts fed sequentially into the model improves action fidelity but compromises temporal consistency, while a single complex prompt preserves consistency at the cost of prompt-following capability. We attribute this problem to two primary causes: 1) temporal misalignment between video content and the prompt, and 2) conflicting attention coupling between motion-related visual objects and their associated text conditions. To address these challenges, we propose a novel, training-free attention mechanism, Temporal-wise Separable Attention (TS-Attn), which dynamically rearranges attention distribution to ensure temporal awareness and global coherence in multi-event scenarios. TS-Attn can be seamlessly integrated into various pre-trained text-to-video models, boosting StoryEval-Bench scores by 33.5% and 16.4% on Wan2.1-T2V-14B and Wan2.2-T2V-A14B with only a 2% increase in inference time. It also supports plug-and-play usage across models for multi-event image-to-video generation. The source code and project page are available at https://github.com/Hong-yu-Zhang/TS-Attn.