FlowAlign: Trajectory-Regularized, Inversion-Free Flow-based Image Editing

2025-05-29

AI Summary

Existing inversion-free text-driven image editing methods (e.g., FlowEdit) suffer from unstable editing trajectories and poor source-image consistency. To address these issues, this paper proposes a novel editing framework built upon pretrained flow-based diffusion models (e.g., Stable Diffusion 3). Our method directly models the invertible noise-to-image flow via ordinary differential equation (ODE) solving, eliminating the need for latent-space inversion. We introduce a flow-matching loss as a trajectory regularization term to jointly optimize semantic alignment and source-structure fidelity. The resulting approach enables fully reversible editing while preserving high source-image fidelity, significantly improving edit controllability and trajectory stability. Extensive experiments demonstrate that our method consistently outperforms state-of-the-art inversion-free approaches across multiple benchmarks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Recent inversion-free, flow-based image editing methods such as FlowEdit leverages a pre-trained noise-to-image flow model such as Stable Diffusion 3, enabling text-driven manipulation by solving an ordinary differential equation (ODE). While the lack of exact latent inversion is a core advantage of these methods, it often results in unstable editing trajectories and poor source consistency. To address this limitation, we propose FlowAlign, a novel inversion-free flow-based framework for consistent image editing with principled trajectory control. FlowAlign introduces a flow-matching loss as a regularization mechanism to promote smoother and more stable trajectories during the editing process. Notably, the flow-matching loss is shown to explicitly balance semantic alignment with the edit prompt and structural consistency with the source image along the trajectory. Furthermore, FlowAlign naturally supports reverse editing by simply reversing the ODE trajectory, highlighting the reversible and consistent nature of the transformation. Extensive experiments demonstrate that FlowAlign outperforms existing methods in both source preservation and editing controllability.

Problem

• Improves unstable editing trajectories in flow-based image editing
• Enhances source consistency during text-driven image manipulation
• Balances semantic alignment and structural consistency in edits

Innovation

• Inversion-free flow-based image editing framework
• Flow-matching loss for stable trajectories
• Reversible ODE trajectory for consistent editing