Blur faces automatically
Finds faces and hides them, without the photo leaving your device.
- Runs on your device — the file is never sent anywhere
- No sign-up, no watermarks, no daily quotas
- Free, with no limit on how often you use it
How it works
Blurring faces is usually about publishing something responsibly — a photo of a crowd, a classroom, a street, a workplace. It is also the clearest case for not uploading: the reason you are blurring is that the faces should not be shared, and an upload shares them.
It runs on your device
The first use downloads the MediaPipe face-detection engine (about 10 MB) from the public jsDelivr CDN, plus a 230 KB model served from this site; your browser then caches both. The photo is never transmitted. Detection takes a moment; hiding the faces is instant.
The pixels are destroyed
Faces are pixelated, blurred or blacked out in the image data itself, not covered by an overlay. Nothing can be recovered from the saved file.
Check the result — this matters
Automatic detection is good on faces that are reasonably large, front-facing and clearly lit. It misses profiles, faces at the edge of the frame, small faces in the background, and faces in shadow. A missed face is precisely the one that matters, so look at the output before publishing. Anything missed can be covered with the manual hide tool.
Area covered
The detected box is expanded slightly by default, because a box drawn tightly around the features often leaves hair, ears and jawline visible — enough for someone to recognise. Increase it if faces are still identifiable.
Blur faces automatically — questions
Is my photo uploaded?
No. The detection model comes to your browser; the photo stays on your device.
Can the blurred faces be recovered?
No. The pixels are destroyed in the image data rather than covered.
It missed a face — why?
Detection struggles with profiles, small background faces and poor lighting. Use the manual hide tool for anything missed, and always check before publishing.
Does it work on video?
No, photos only.