Multi-Subject Personalization

TL;DR

This paper introduces Multi-Subject Personalization (MSP), a method to improve the consistency and quality of images with multiple personalized subjects in text-to-image generation, addressing existing challenges in rendering interactions.

Contribution

The paper proposes MSP, a novel approach using Stable Diffusion to enhance multi-subject personalization in text-to-image models, improving coherence and quality.

Findings

MSP produces more consistent multi-subject images.

MSP outperforms other models in rendering interactions.

MSP maintains high image quality across subjects.

Abstract

Creative story illustration requires a consistent interplay of multiple characters or objects. However, conventional text-to-image models face significant challenges while producing images featuring multiple personalized subjects. For example, they distort the subject rendering, or the text descriptions fail to render coherent subject interactions. We present Multi-Subject Personalization (MSP) to alleviate some of these challenges. We implement MSP using Stable Diffusion and assess our approach against other text-to-image models, showcasing its consistent generation of good-quality images representing intended subjects and interactions.