How to blur faces in a photo?

1.
Add a photo
Drop a JPG, PNG or WebP file or paste an image with Ctrl+V. On first use the face detection model is downloaded (about 4 MB).
2.
Check detected faces
Faces are marked automatically with an ellipse. Drag across the photo to add an area - a rectangle, an ellipse or a freehand shape. Click an area to move it, resize it with the corner handle or remove it with the × button or the Delete key.
3.
Choose the effect
Blur, pixelate or black out, and the effect strength. The effect covers exactly the shape of each area and the preview updates live.
4.
Download
Click “Download photo” and choose JPG, PNG or WebP. The output file contains no EXIF metadata.

When should faces be hidden?

Under the GDPR a person’s image is personal data, and publishing it usually requires the consent of the person shown. Hiding faces is a simple way to share photos from school or company events, damage reports, CCTV footage, property listings or training materials without exposing the people in them. Exceptions such as crowds at public events depend on local law, so when in doubt, anonymise.

How does automatic face detection work?

The tool uses the BlazeFace model from Google MediaPipe (Apache-2.0), running in your browser on the graphics card (WebGL) or the processor. Because the model sees a low-resolution image, the “Also look for small faces” option additionally scans overlapping tiles to find faces in the background and in group photos. Detected boxes are enlarged to cover the forehead, hair and chin, and repeated detections of the same face from different tiles, as well as single eyes mistaken for faces, are merged into one area.

Only the model (about 0.23 MB, from Google servers) and the MediaPipe runtime (about 3.5 MB, from the jsDelivr CDN) are downloaded, once. The photo itself is never uploaded.

Blur, pixelate or black out?

Effect	Look	Safety
Blur	most natural	good at strength 50-100
Pixelate	clearly censored	good with large blocks
Black out	black ellipse, rectangle or any shape	highest - the information is removed

Weak blur or fine pixelation may still let people recognise the person. For sensitive images (children, CCTV) use black out or a strength above 70. Strength is relative to face size, so small and large faces are hidden equally well.

Limitations and good practice

Always check the result - faces in profile, turned away, covered or very small can be missed. Add them manually.
Not only faces - tattoos, licence plates or name badges can identify a person; cover them manually.
Large photos - images above 6000 px on the longer side are scaled down to avoid running out of memory.

Frequently asked questions

Is my photo uploaded to a server?

No. Your browser detects and blurs the faces. Only the MediaPipe model and runtime are downloaded (about 4 MB, once).

What if a face was not detected?

Drag across the photo to mark the area manually. Turning on “Also look for small faces” also helps.

Can the blur be reversed?

Black out cannot be reversed. Strong blur and large pixelation are practically irreversible, but weak effects may still allow recognition - use strength above 70 or black out.

Does it work on a phone?

Yes, in current mobile browsers. Add and move areas with your finger and trace freehand shapes like with a marker.

Can I blur licence plates?

Yes - mark them manually and choose the rectangle shape. Automatic detection works for faces only.

Updated: October 11, 2026