Explore current models
English

Codeformer / codeformer face restoration

Restore faces in older photographs and AI-generated portraits.

Browse current models on Synexa Status: At latest check, Synexa CodeFormer model page returned an error; explore current offerings on the Synexa catalog instead.
About this model

How Face Restoration Models Work

CodeFormer and similar models accept an image file as input, along with configurable settings, rather than a text prompt.

Quality vs. Fidelity (0 to 1)

Balancing enhancement (lower values) against preserving original facial features (higher values).

Background Enhancement (Boolean)

Optional toggle to process the rest of the image using Real-ESRGAN.

Face Upsampling (Boolean)

Toggle upsampling for high-resolution output scaling.

Important Status Notice

This is an independent editorial guide. We do not host, run, or provide direct access to the CodeFormer model here. At our latest check, the CodeFormer model page on Synexa returned an error; we link to the Synexa homepage to browse current offerings instead.

You can explore alternative models or check current availability directly at Synexa.

Explore Models on Synexa

Examples (Conceptual Illustrations - Not Model Outputs)

A faded family portrait with closely grouped faces and the warmth of an old print.
A softly lit classic studio portrait framed against a dark background.
A detailed fantasy portrait with ornate clothing and a leafy backdrop.
A candid photograph with a face in gentle focus and warm natural light.
An atmospheric portrait in a shaded garden with delicate light falling across the face.

Run time and cost

Run time and cost depend on the image, selected model, and settings. Check the destination platform for the options available before starting a run.

CodeFormer is a face restoration approach for photographs and AI-generated faces. Larger images and additional enhancement steps may take longer to process.

Readme

Codeformer

Towards Robust Blind Face Restoration with Codebook Lookup Transformer (NeurIPS 2022) · Research paper · Project overview

CodeFormer research explores restoring facial detail when the original image is degraded. This page provides an editorial overview and conceptual illustrations of the approach, not a hosted image-restoration tool.

☕ Face restoration for old photos and AI-generated faces

The quality–fidelity trade-off matters: a stronger restoration may produce a clearer face, while a more faithful setting aims to preserve more of the source image.

For research or implementation details, consult the original CodeFormer work and its license before using the model commercially.

License · The original CodeFormer project uses the S-Lab License 1.0. Review its terms for permitted uses, particularly non-commercial use.

Citation · If the research is useful to your work, cite the original paper:

@inproceedings{zhou2022codeformer,
  author = {Zhou, Shangchen and Chan, Kelvin C.K. and Li, Chongyi and Loy, Chen Change},
  title = {Towards Robust Blind Face Restoration with Codebook Lookup Transformer},
  booktitle = {NeurIPS},
  year = {2022}
}
Explore current models
Explore current models