TempoMaster: Efficient Long Video Generation via Next-Frame-Rate Prediction

TL;DR

TempoMaster introduces a novel method for long video generation by predicting and refining frame rates, achieving high-quality, temporally coherent videos efficiently through a multi-stage process.

Contribution

It proposes a new framework that generates long videos by progressively increasing frame rates, combining bidirectional attention and autoregression for improved coherence and efficiency.

Findings

Sets new state-of-the-art in long video generation

Achieves superior visual and temporal quality

Enables efficient parallel synthesis of videos

Abstract

We present TempoMaster, a novel framework that formulates long video generation as next-frame-rate prediction. Specifically, we first generate a low-frame-rate clip that serves as a coarse blueprint of the entire video sequence, and then progressively increase the frame rate to refine visual details and motion continuity. During generation, TempoMaster employs bidirectional attention within each frame-rate level while performing autoregression across frame rates, thus achieving long-range temporal coherence while enabling efficient and parallel synthesis. Extensive experiments demonstrate that TempoMaster establishes a new state-of-the-art in long video generation, excelling in both visual and temporal quality.