Free AI Face Enhancer
AI Face Enhancer restores blurry, old or low-resolution faces: it finds up to 20 per photo and rebuilds each one with GFPGAN, replacing only the area around each face. Lower the strength to keep more of the original look, untick faces to leave them as they were, and save at the original size or 2×.
- Faces:
- Up to 20 per photo
- Models:
- YuNet and GFPGAN v1.4
- Output:
- Original size or 2×
- Price:
- Free
- Runs:
- AI model in your browser
- Uploads:
- None: files never leave your device
- Account:
- Not needed
Quick answers
Are my photos uploaded?
No. Face detection and restoration run in your browser; only the AI models are downloaded once from Hugging Face. Your photos never leave your device.
Does it change the rest of the photo?
No. Only the area around each face is replaced, with a soft edge. Everything else stays as it was (or is simply enlarged if you choose 2×).
Processed on your device: your files are never uploaded. How it works
Network activity in this tab
What this page has requested since you opened the site in this tab. An upload is any request that sends data: POST, PUT, PATCH or a request with a body.
No uploads: nothing has been sent from this page.
No network requests yet.
To see every request, open your browser's developer tools (Network tab).
How to use the AI Face Enhancer
- Drop a photo with one or more faces (old, blurry, small or compressed).
- Click Enhance faces: the first time the AI models (about 170 MB) are downloaded and cached.
- Compare before and after, adjust the restoration strength and untick faces you want to keep as they were.
- Choose the original size or 2×, then download the result.
Credits & open source
Open-source AI model running in your browser. Thank you to the people who make these projects.
- ONNX Runtime WebExecutes ONNX models with WebGPU or WebAssembly (self-hosted from /vendor/ort). © Microsoft Corporation · MIT
- picaHigh-quality image resizing (Lanczos filter) in the browser. © Vitaly Puzrin · MIT
Web platform APIs used: Web Workers, WebAssembly, WebGPU, Canvas 2D, createImageBitmap, File API.
- GFPGAN v1.4 (ONNX, fp16-stored weights) by Xintao Wang, Yu Li, Honglun Zhang, Ying Shan (Tencent ARC Lab); ONNX conversion by opiumuseless from Meeperomi/GFPGANv1.4-onnx (Apache-2.0) — AI model that restores each face
- YuNet face detector (2023mar ONNX) by Wei Wu, Hanyang Peng, Shiqi Yu (OpenCV Zoo) (MIT) — finds faces and their five landmarks
Full license texts: third-party notices · all credits
Frequently asked questions
Will the restored face look exactly like the person?
The model rebuilds missing details such as eyes, skin and teeth, so very damaged faces may look slightly different. Lower the strength to keep more of the original.
Why was a face not found?
Faces seen from the side, partly hidden or only a few pixels wide may be missed. For a close-up portrait you can restore the whole picture as one face.
Is it slow on my computer?
Each face takes a second or two with WebGPU and around 10–30 seconds on the processor. Recent Chrome and Edge use the graphics card.