PhyCo: Learning Controllable Physical Priors for Generative Motion

TL;DR

PhyCo is a novel framework that enhances video diffusion models with controllable, physically consistent generation by integrating a large simulation dataset, physics-supervised fine-tuning, and vision-language guided optimization.

Contribution

It introduces a scalable method combining simulation data, physics supervision, and visual language feedback to produce physically realistic and controllable videos without requiring geometry reconstruction.

Findings

PhyCo outperforms baselines on the Physics-IQ benchmark.

Human studies show improved physical realism and control.

The approach generalizes beyond synthetic training environments.

Abstract

Modern video diffusion models excel at appearance synthesis but still struggle with physical consistency: objects drift, collisions lack realistic rebound, and material responses seldom match their underlying properties. We present PhyCo, a framework that introduces continuous, interpretable, and physically grounded control into video generation. Our approach integrates three key components: (i) a large-scale dataset of over 100K photorealistic simulation videos where friction, restitution, deformation, and force are systematically varied across diverse scenarios; (ii) physics-supervised fine-tuning of a pretrained diffusion model using a ControlNet conditioned on pixel-aligned physical property maps; and (iii) VLM-guided reward optimization, where a fine-tuned vision-language model evaluates generated videos with targeted physics queries and provides differentiable feedback. This combination enables a generative model to produce physically consistent and controllable outputs through variations in physical attributes-without any simulator or geometry reconstruction at inference. On the Physics-IQ benchmark, PhyCo significantly improves physical realism over strong baselines, and human studies confirm clearer and more faithful control over physical attributes. Our results demonstrate a scalable path toward physically consistent, controllable generative video models that generalize beyond synthetic training environments.