CodeFormer

Name: CodeFormer
Author: Shangchen Zhou, S-Lab, Nanyang Technological University

AI face restoration tool that repairs blurry or degraded faces in photos and videos.

ProprietaryBy Shangchen Zhou, S-Lab, Nanyang Technological University

FreeProprietaryWindows macOS Linux Command line

Official websitegithub.com

Download pagegithub.com

Compare CodeFormer with another product

Alternatives

GFPGANGFPGAN is another locally run face restoration tool from Tencent ARC with multiple model versions and optional Real-ESRGAN background enhancement, but it lacks CodeFormer's video input and colorization scripts.
Restore PhotosRestore Photos is a single-purpose web app for restoring old and blurry face photos with a side-by-side view, and its source code is published on GitHub.
ImageColorizerAn online AI service that colorizes black-and-white photos and repairs old, damaged pictures.
Real-ESRGANOpen source AI model for upscaling and restoring images and video.
MyHeritage Photo EnhancerMyHeritage Photo Enhancer sharpens faces in the browser with nothing to install, but it is tied to the MyHeritage genealogy service and does not take video input.
ReminiRemini replaces the Python and CUDA setup with one-step face enhancement in the browser and a mobile app, but photos are processed online.

See all alternatives

About CodeFormer

CodeFormer is a face restoration model from S-Lab at Nanyang Technological University, presented at NeurIPS 2022. It uses a codebook lookup transformer to rebuild faces in blurry, damaged or low-quality images, and also works on AI-generated art and video files.

Beyond restoration it includes scripts for face colorization and face inpainting on cropped and aligned images. It runs locally with PyTorch and CUDA from Python scripts, and online demos on Hugging Face, Replicate and OpenXLab let you try it without a setup.

Strengths

Restores heavily degraded faces
Accepts video input as well as images
Extra scripts for face colorization and inpainting
Training code is published

Limitations

Needs Python, PyTorch and a CUDA setup
License terms are not stated clearly on the GitHub page
Command-line scripts rather than a graphical app

Details

Pricing: FreeFree to download and run from the GitHub repository.
License: Proprietary
Developer: Shangchen Zhou, S-Lab, Nanyang Technological University
Platforms: Windows, macOS, Linux, Command line
How it runs: Downloadable app
Account: Not required
Works offline: Yes
Best suited for: Fixing blurry faces in old photos, video frames or AI art on your own machine
Categories: Local AI tools, AI photo editing
Last verified: October 2, 2026
Added: October 2, 2026
Sources: github.com/sczhou/CodeFormer

Report a wrong fact or a dead link on this listing