Temporal Regularization Makes Your Video Generator Stronger
This work addresses the problem of temporal quality in video generation for AI researchers, offering a simple, model-agnostic method that is incremental but effective.
The paper tackles the challenge of achieving high temporal coherence and diversity in video generation by introducing FluxFlow, a temporal augmentation strategy that applies controlled perturbations at the data level. Results show significant improvements in temporal coherence and diversity across models like U-Net, DiT, and AR-based architectures on UCF-101 and VBench benchmarks, while preserving spatial fidelity.
Temporal quality is a critical aspect of video generation, as it ensures consistent motion and realistic dynamics across frames. However, achieving high temporal coherence and diversity remains challenging. In this work, we explore temporal augmentation in video generation for the first time, and introduce FluxFlow for initial investigation, a strategy designed to enhance temporal quality. Operating at the data level, FluxFlow applies controlled temporal perturbations without requiring architectural modifications. Extensive experiments on UCF-101 and VBench benchmarks demonstrate that FluxFlow significantly improves temporal coherence and diversity across various video generation models, including U-Net, DiT, and AR-based architectures, while preserving spatial fidelity. These findings highlight the potential of temporal augmentation as a simple yet effective approach to advancing video generation quality.