ConceptWeaver: Weaving Disentangled Concepts with Flow

TL;DR

ConceptWeaver introduces a stage-aware framework for disentangling and manipulating concepts in flow-based models, enabling precise, high-fidelity content editing from a single reference image.

Contribution

It reveals the three-stage generative process of flow models and leverages this insight to develop a novel method for one-shot concept disentanglement and editing.

Findings

Enables high-fidelity, compositional synthesis and editing.

Validates the staged approach improves concept manipulation.

Demonstrates effectiveness across diverse scenarios.

Abstract

Pre-trained flow-based models excel at synthesizing complex scenes yet lack a direct mechanism for disentangling and customizing their underlying concepts from one-shot real-world sources. To demystify this process, we first introduce a novel differential probing technique to isolate and analyze the influence of individual concept tokens on the velocity field over time. This investigation yields a critical insight: the generative process is not monolithic but unfolds in three distinct stages. An initial \textbf{Blueprint Stage} establishes low-frequency structure, followed by a pivotal \textbf{Instantiation Stage} where content concepts emerge with peak intensity and become naturally disentangled, creating an optimal window for manipulation. A final concept-insensitive refinement stage then synthesizes fine-grained details. Guided by this discovery, we propose \textbf{ConceptWeaver}, a framework for one-shot concept disentanglement. ConceptWeaver learns concept-specific semantic offsets from a single reference image using a stage-aware optimization strategy that aligns with the three-stage framework. These learned offsets are then deployed during inference via our novel ConceptWeaver Guidance (CWG) mechanism, which strategically injects them at the appropriate generative stage. Extensive experiments validate that ConceptWeaver enables high-fidelity, compositional synthesis and editing, demonstrating that understanding and leveraging the intrinsic, staged nature of flow models is key to unlocking precise, multi-granularity content manipulation.