FreeLoRA: Enabling Training-Free LoRA Fusion for Autoregressive Multi-Subject Personalization
This addresses the challenge of generating images with multiple personalized subjects for applications like virtual try-on, though it is incremental as it builds on LoRA-based adaptations.
The paper tackles the problem of multi-subject personalization in image generation, where existing methods require complex re-tuning, and presents FreeLoRA, a framework that enables training-free fusion of subject-specific LoRA modules, achieving strong performance in subject fidelity and prompt consistency.
Subject-driven image generation plays a crucial role in applications such as virtual try-on and poster design. Existing approaches typically fine-tune pretrained generative models or apply LoRA-based adaptations for individual subjects. However, these methods struggle with multi-subject personalization, as combining independently adapted modules often requires complex re-tuning or joint optimization. We present FreeLoRA, a simple and generalizable framework that enables training-free fusion of subject-specific LoRA modules for multi-subject personalization. Each LoRA module is adapted on a few images of a specific subject using a Full Token Tuning strategy, where it is applied across all tokens in the prompt to encourage weakly supervised token-content alignment. At inference, we adopt Subject-Aware Inference, activating each module only on its corresponding subject tokens. This enables training-free fusion of multiple personalized subjects within a single image, while mitigating overfitting and mutual interference between subjects. Extensive experiments show that FreeLoRA achieves strong performance in both subject fidelity and prompt consistency.