GFPGAN

GFPGAN (Generative Facial Prior GAN) is a powerful open-source face restoration tool developed by TencentARC. With over 37,000 GitHub stars, it uses a GAN architecture to restore and enhance facial details in low-resolution, blurry, or damaged photos. GFPGAN combines a pre-trained face GAN as a prior with a U-Net-based restoring network, leveraging the rich and diverse priors encapsulated in a pretrained face GAN (StyleGAN) for blind face restoration. It can fix blurry faces, remove artifacts, enhance eyes and facial features, and produce photo-realistic results. The tool supports standalone image processing, batch enhancement, real-time video face restoration, and can be integrated with other restoration tools like Real-ESRGAN for complete photo enhancement. It runs on consumer GPUs with as little as 2GB VRAM and also supports CPU-only operation.