TencentARC/GFPGAN

Project overview

GFPGAN is a Python library and CLI tool for real-world blind face restoration, leveraging generative priors from a pretrained face GAN. It is operated locally via command-line inference, accepts image or folder inputs, and outputs restored images.

Project type

Model Runtime · Image & Vision

Use cases

Image Processing

Deployment

Refer to project documentation

License

License pending

Best for

• Developers and AI engineers who need a locally runnable library and CLI for blind face restoration using pretrained GAN models.
• Users who require training codes to customize or improve face restoration models.

Key capabilities

• Performs blind face restoration using generative facial priors from a pretrained face GAN.
• Provides training codes to improve or customize the model.

Limitations and risks

• The V1.3 model is not very sharp and has a slight change on identity.
• The V1.2 model produces some unnatural outputs.

Getting started

• Clone the repository, install dependent packages including basicsr, facexlib, and the realesrgan package, then run python inference_gfpgan.py. The setup difficulty is medium.

Evidence and sources

• GitHub project description: GFPGAN aims at developing Practical Algorithms for Real-world Face Restoration.
• README: GFPGAN aims at developing a **Practical Algorithm for Real-world Face Restoration**.<br> It leverages rich and diverse priors encapsulated in a pretrained face GAN (*e.g.*, StyleG…
• README: - :white_check_mark: Add [RestoreFormer](https://github.com/wzhouxiff/RestoreFormer) inference codes. - :white_check_mark: Add [V1.4 model](https://github.com/TencentARC/GFPGAN/re…
• README: python inference_gfpgan.py -i inputs/whole_imgs -o results -v 1.3 -s 2 ``` ```console Usage: python inference_gfpgan.py -i inputs/whole_imgs -o results -v 1.3 -s 2 [options]... -h…
• README: Download pre-trained models: [GFPGANv1.3.pth](https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth)

Reviewed at: 2026-07-17T10:07:28Z