Constrained Edit Fields for Training-Free Flow Editing

2026-09-27

AI Summary

This study addresses the issue of unintended alterations in non-target regions caused by trajectory accumulation during training-free flow editing. To mitigate this, we propose a constrained editing field method that introduces, for the first time, a continuous-space editing responsibility mechanism. Specifically, the base editing field is decomposed into responsibility components attributed to the source image and the proposal, while an unconstrained proposal estimates missing content to support edited regions. This formulation preserves instruction-relevant updates while suppressing unexpected changes, achieving trajectory modification atop pretrained rectified flow models. Experiments demonstrate that our approach attains state-of-the-art performance in structural distance and background preservation on PIE-Bench. Under Stable Diffusion 3.5, it reduces background mean squared error by 48% while maintaining high instruction alignment.

Abstract

Text-guided image editing aims to perform a desired edit while preserving source content unrelated to it. Pretrained rectified-flow models enable training-free editing of real images through modifications to their sampling trajectories. However, responses at locations unrelated to the desired edit can still accumulate along the editing trajectory and become visible in the final result. To overcome this, we propose Constrained Edit Fields (CEF), which assigns each spatial location a continuous edit responsibility that quantifies its relevance to the desired edit. CEF estimates edit responsibility directly from the source image when the relevant content is present. For edits whose target content is absent from the source, CEF first generates an unconstrained proposal to reveal its realized spatial support and then estimates responsibility from that proposal. At each editing step, CEF decomposes the base edit field into prompt-induced and trajectory-induced components, enabling edit responsibility to preserve instruction-relevant updates while suppressing unintended trajectory-induced changes. Evaluated on all 700 PIE-Bench examples, CEF achieves state-of-the-art Structure Distance, background LPIPS, and background MSE with both Stable Diffusion 3.5 Medium and FLUX, while retaining competitive instruction alignment. On Stable Diffusion 3.5 Medium, it reduces these metrics over the previous best results by 10.2%, 21.2%, and 48.0%, respectively.

Problem

• text-guided image editing
• training-free editing
• rectified-flow models
• unintended changes
• content preservation

Innovation

• Training-Free Flow Editing
• Constrained Edit Fields
• Rectified-Flow Models
• Edit Responsibility
• Text-Guided Image Editing