Uniform Discrete Diffusion with Metric Path for Video Generation

TL;DR

URSA introduces a novel discrete diffusion framework with metric path and resolution-dependent timestep shifting, enabling efficient high-resolution and long-duration video generation with fewer inference steps, matching continuous methods' performance.

Contribution

It presents URSA, a new discrete video generation method with innovative designs that improve scalability and efficiency, bridging the gap with continuous approaches.

Findings

Outperforms existing discrete methods on benchmarks.

Achieves comparable performance to state-of-the-art continuous diffusion models.

Supports versatile tasks like interpolation and image-to-video generation.

Abstract

Continuous-space video generation has advanced rapidly, while discrete approaches lag behind due to error accumulation and long-context inconsistency. In this work, we revisit discrete generative modeling and present Uniform discRete diffuSion with metric pAth (URSA), a simple yet powerful framework that bridges the gap with continuous approaches for the scalable video generation. At its core, URSA formulates the video generation task as an iterative global refinement of discrete spatiotemporal tokens. It integrates two key designs: a Linearized Metric Path and a Resolution-dependent Timestep Shifting mechanism. These designs enable URSA to scale efficiently to high-resolution image synthesis and long-duration video generation, while requiring significantly fewer inference steps. Additionally, we introduce an asynchronous temporal fine-tuning strategy that unifies versatile tasks within a single model, including interpolation and image-to-video generation. Extensive experiments on challenging video and image generation benchmarks demonstrate that URSA consistently outperforms existing discrete methods and achieves performance comparable to state-of-the-art continuous diffusion methods. Code and models are available at https://github.com/baaivision/URSA