Overview
GFPGAN is a research implementation for blind real-world face restoration. Its model uses priors from a pretrained face GAN to recover facial details in blurry, compressed, or otherwise degraded portraits, and the repository includes inference scripts and pretrained model options. The project is intended for developers and researchers who run the code locally with Python and PyTorch. It can also enhance non-face background areas through Real-ESRGAN integration. GitHub reports no SPDX license assertion, so this record does not classify it as an OSI-licensed open-source product under the directory rules.