ProEdit: Inversion-based Editing From Prompts Done Right

TL;DR

ProEdit introduces novel attention and latent techniques to improve inversion-based visual editing, achieving state-of-the-art results while maintaining background consistency and allowing seamless integration with existing methods.

Contribution

The paper proposes KV-mix and Latents-Shift to address source influence issues in inversion-based editing, enhancing editing accuracy and consistency.

Findings

Achieves state-of-the-art performance on image and video editing benchmarks.

The methods are plug-and-play and compatible with existing inversion techniques.

Significantly improves editing attribute changes without compromising background quality.

Abstract

Inversion-based visual editing provides an effective and training-free way to edit an image or a video based on user instructions. Existing methods typically inject source image information during the sampling process to maintain editing consistency. However, this sampling strategy overly relies on source information, which negatively affects the edits in the target image (e.g., failing to change the subject's atributes like pose, number, or color as instructed). In this work, we propose ProEdit to address this issue both in the attention and the latent aspects. In the attention aspect, we introduce KV-mix, which mixes KV features of the source and the target in the edited region, mitigating the influence of the source image on the editing region while maintaining background consistency. In the latent aspect, we propose Latents-Shift, which perturbs the edited region of the source latent, eliminating the influence of the inverted latent on the sampling. Extensive experiments on several image and video editing benchmarks demonstrate that our method achieves SOTA performance. In addition, our design is plug-and-play, which can be seamlessly integrated into existing inversion and editing methods, such as RF-Solver, FireFlow and UniEdit.