DiffFAE introduces a one-stage diffusion-based framework for high-fidelity facial appearance editing, addressing challenges of low generation fidelity, poor attribute preservation, and inefficient inference.
