RDPO: Real Data Preference Optimization for Physics Consistency Video Generation

TL;DR

RDPO is an annotation-free framework that enhances physical realism in video generation by distilling physical priors directly from real videos through iterative training, improving action coherence and realism.

Contribution

RDPO introduces a novel method to incorporate physical priors into video generation without human annotations, using real videos to guide the generator via preference pairs.

Findings

Significantly improves physical realism in generated videos.

Enhances action coherence and physical consistency.

Achieves state-of-the-art results on multiple benchmarks.

Abstract

Video generation techniques have achieved remarkable advancements in visual quality, yet faithfully reproducing real-world physics remains elusive. Preference-based model post-training may improve physical consistency, but requires costly human-annotated datasets or reward models that are not yet feasible. To address these challenges, we present Real Data Preference Optimisation (RDPO), an annotation-free framework that distills physical priors directly from real-world videos. Specifically, the proposed RDPO reverse-samples real video sequences with a pre-trained generator to automatically build preference pairs that are statistically distinguishable in terms of physical correctness. A multi-stage iterative training schedule then guides the generator to obey physical laws increasingly well. Benefiting from the dynamic information explored from real videos, our proposed RDPO significantly improves the action coherence and physical realism of the generated videos. Evaluations on multiple benchmarks and human evaluations have demonstrated that RDPO achieves improvements across multiple dimensions. The source code and demonstration of this paper are available at: https://wwenxu.github.io/RDPO/