Semantic Editing with Coupled Stochastic Differential Equations
This addresses the problem of precise image editing for users of generative AI, offering an incremental improvement over existing methods.
The paper tackles the challenge of editing images with pretrained text-to-image models, which often distort details or cause artifacts, by proposing coupled stochastic differential equations (coupled SDEs) to guide sampling, achieving high prompt fidelity and near-pixel-level consistency without retraining.
Editing the content of an image with a pretrained text-to-image model remains challenging. Existing methods often distort fine details or introduce unintended artifacts. We propose using coupled stochastic differential equations (coupled SDEs) to guide the sampling process of any pre-trained generative model that can be sampled by solving an SDE, including diffusion and rectified flow models. By driving both the source image and the edited image with the same correlated noise, our approach steers new samples toward the desired semantics while preserving visual similarity to the source. The method works out-of-the-box-without retraining or auxiliary networks-and achieves high prompt fidelity along with near-pixel-level consistency. These results position coupled SDEs as a simple yet powerful tool for controlled generative AI.