From Statics to Dynamics: Physics-Aware Image Editing with Latent Transition Priors

TL;DR

This paper introduces PhysicEdit, a physics-aware image editing framework that models physical state transitions, leveraging a large video dataset and a dual-thinking reasoning mechanism to produce more physically plausible edits.

Contribution

It formulates image editing as predictive physical state transitions and introduces PhysicTran38K dataset and PhysicEdit framework with a dual-thinking mechanism for improved realism.

Findings

PhysicEdit outperforms previous models in physical realism by 5.9%.

PhysicEdit improves knowledge-grounded editing accuracy by 10.1%.

Achieves state-of-the-art results among open-source methods.

Abstract

Instruction-based image editing has achieved remarkable success in semantic alignment, yet state-of-the-art models frequently fail to render physically plausible results when editing involves complex causal dynamics, such as refraction or material deformation. We attribute this limitation to the dominant paradigm that treats editing as a discrete mapping between image pairs, which provides only boundary conditions and leaves transition dynamics underspecified. To address this, we reformulate physics-aware editing as predictive physical state transitions and introduce PhysicTran38K, a large-scale video-based dataset comprising 38K transition trajectories across five physical domains, constructed via a two-stage filtering and constraint-aware annotation pipeline. Building on this supervision, we propose PhysicEdit, an end-to-end framework equipped with a textual-visual dual-thinking mechanism. It combines a frozen Qwen2.5-VL for physically grounded reasoning with learnable transition queries that provide timestep-adaptive visual guidance to a diffusion backbone. Experiments show that PhysicEdit improves over Qwen-Image-Edit by 5.9% in physical realism and 10.1% in knowledge-grounded editing, setting a new state-of-the-art for open-source methods, while remaining competitive with leading proprietary models.