LoRACLR: Contrastive Adaptation for Customization of Diffusion Models

TL;DR

LoRACLR introduces a contrastive adaptation method that effectively merges multiple personalized diffusion models into a single model, enabling high-quality multi-concept image generation without additional fine-tuning.

Contribution

The paper proposes a novel contrastive learning approach to combine multiple LoRA models into one, maintaining concept distinctiveness and compatibility without extra training.

Findings

Successfully merges multiple concepts into a single model

Enables high-quality multi-concept image synthesis

Reduces need for separate fine-tuning

Abstract

Recent advances in text-to-image customization have enabled high-fidelity, context-rich generation of personalized images, allowing specific concepts to appear in a variety of scenarios. However, current methods struggle with combining multiple personalized models, often leading to attribute entanglement or requiring separate training to preserve concept distinctiveness. We present LoRACLR, a novel approach for multi-concept image generation that merges multiple LoRA models, each fine-tuned for a distinct concept, into a single, unified model without additional individual fine-tuning. LoRACLR uses a contrastive objective to align and merge the weight spaces of these models, ensuring compatibility while minimizing interference. By enforcing distinct yet cohesive representations for each concept, LoRACLR enables efficient, scalable model composition for high-quality, multi-concept image synthesis. Our results highlight the effectiveness of LoRACLR in accurately merging multiple concepts, advancing the capabilities of personalized image generation.