UltraGen: High-Resolution Video Generation with Hierarchical Attention

TL;DR

UltraGen introduces a hierarchical attention framework that enables efficient, high-resolution video generation up to 4K, overcoming previous computational limitations of diffusion transformer models.

Contribution

It presents a novel hierarchical dual-branch attention architecture with global-local decomposition and spatial compression for scalable high-resolution video synthesis.

Findings

Successfully scales models to 1080P and 4K resolutions.

Outperforms existing state-of-the-art methods in quality and efficiency.

Enables end-to-end native high-resolution video generation.

Abstract

Recent advances in video generation have made it possible to produce visually compelling videos, with wide-ranging applications in content creation, entertainment, and virtual reality. However, most existing diffusion transformer based video generation models are limited to low-resolution outputs (<=720P) due to the quadratic computational complexity of the attention mechanism with respect to the output width and height. This computational bottleneck makes native high-resolution video generation (1080P/2K/4K) impractical for both training and inference. To address this challenge, we present UltraGen, a novel video generation framework that enables i) efficient and ii) end-to-end native high-resolution video synthesis. Specifically, UltraGen features a hierarchical dual-branch attention architecture based on global-local attention decomposition, which decouples full attention into a local attention branch for high-fidelity regional content and a global attention branch for overall semantic consistency. We further propose a spatially compressed global modeling strategy to efficiently learn global dependencies, and a hierarchical cross-window local attention mechanism to reduce computational costs while enhancing information flow across different local windows. Extensive experiments demonstrate that UltraGen can effectively scale pre-trained low-resolution video models to 1080P and even 4K resolution for the first time, outperforming existing state-of-the-art methods and super-resolution based two-stage pipelines in both qualitative and quantitative evaluations.