HyperLoRA: Parameter-Efficient Adaptive Generation for Portrait Synthesis

Wizard hat Pink wavy hair Cyberpunk Snow Chinese clothing Beach HyperLoRA w.o. fine-tuning Figure 1. We propose HyperLoRA, a parameter-efficient adaptive method for portrait synthesis. Given an input face image, HyperLoRA generates personalized LoRA weights without online fine-tuning. Due to the natural interpolability of LoRA, it is easy to support multiple inputs by simple averaging. Leveraging the generated LoRA, we can create personalized portraits with high photorealism and fidelity.