MoA: Mixture-of-Attention for Subject-Context Disentanglement in Personalized Image Generation

TL;DR

MoA introduces a novel Mixture-of-Attention architecture for personalized image generation, enabling disentangled control over subjects and context by combining fixed prior attention with learned personalization pathways.

Contribution

This work presents a new Mixture-of-Attention mechanism that improves subject-context disentanglement in personalized text-to-image diffusion models.

Findings

MoA achieves high-quality personalized images with diverse subject interactions.

The routing mechanism effectively blends personalized and prior content.

MoA enhances control over subject and context separation.

Abstract

We introduce a new architecture for personalization of text-to-image diffusion models, coined Mixture-of-Attention (MoA). Inspired by the Mixture-of-Experts mechanism utilized in large language models (LLMs), MoA distributes the generation workload between two attention pathways: a personalized branch and a non-personalized prior branch. MoA is designed to retain the original model's prior by fixing its attention layers in the prior branch, while minimally intervening in the generation process with the personalized branch that learns to embed subjects in the layout and context generated by the prior branch. A novel routing mechanism manages the distribution of pixels in each layer across these branches to optimize the blend of personalized and generic content creation. Once trained, MoA facilitates the creation of high-quality, personalized images featuring multiple subjects with compositions and interactions as diverse as those generated by the original model. Crucially, MoA enhances the distinction between the model's pre-existing capability and the newly augmented personalized intervention, thereby offering a more disentangled subject-context control that was previously unattainable. Project page: https://snap-research.github.io/mixture-of-attention