HonestFace: Towards Honest Face Restoration with One-Step Diffusion Model
HonestFace proposes a novel one-step diffusion model for face restoration that focuses on generating 'honest' results—faces with true identity and realistic textures, avoiding unnatural over-smoothing. It achieves this by integrating an identity embedder to preserve likeness and a masked face alignment method to enhance local details. Tested on a new real-world dataset, it surpasses existing methods in both visual quality and quantitative metrics.