AnyMS: Bottom-up Attention Decoupling for Layout-guided and Training-free Multi-subject Customization

TL;DR

AnyMS is a training-free framework that uses bottom-up attention decoupling to effectively synthesize multi-subject images guided by layout, text, and subject images, balancing alignment, identity, and layout control.

Contribution

It introduces a novel bottom-up dual-level attention decoupling mechanism for layout-guided multi-subject customization without additional training.

Findings

Achieves state-of-the-art performance in multi-subject image synthesis.

Supports complex compositions with multiple subjects.

Operates without subject-specific training or tuning.

Abstract

Multi-subject customization aims to synthesize multiple user-specified subjects into a coherent image. To address issues such as subjects missing or conflicts, recent works incorporate layout guidance to provide explicit spatial constraints. However, existing methods still struggle to balance three critical objectives: text alignment, subject identity preservation, and layout control, while the reliance on additional training further limits their scalability and efficiency. In this paper, we present AnyMS, a novel training-free framework for layout-guided multi-subject customization. AnyMS leverages three input conditions: text prompt, subject images, and layout constraints, and introduces a bottom-up dual-level attention decoupling mechanism to harmonize their integration during generation. Specifically, global decoupling separates cross-attention between textual and visual conditions to ensure text alignment. Local decoupling confines each subject's attention to its designated area, which prevents subject conflicts and thus guarantees identity preservation and layout control. Moreover, AnyMS employs pre-trained image adapters to extract subject-specific features aligned with the diffusion model, removing the need for subject learning or adapter tuning. Extensive experiments demonstrate that AnyMS achieves state-of-the-art performance, supporting complex compositions and scaling to a larger number of subjects.