Blur a face in a photo

The photo is sent to our servers over an encrypted connection and examined by YuNet, a small open-source face detector published by the OpenCV project. It answers one question, where in this picture is something shaped like a face, and returns a rectangle for each. It does not recognise who anyone is, does not compare faces with any database, and keeps nothing.

Large pictures are also examined in overlapping tiles, because a face that is only a few pixels wide after the whole photo is scaled down would otherwise be overlooked. Very close portraits are checked at a reduced size for the opposite reason.

A light blur softens a face but leaves it recognisable to anyone who knows the person: the outline of the hair, the glasses and the skin tone all survive. If the aim is that nobody can tell who it is, push the slider well past the middle until the features merge into a smooth patch. If the aim is only to make a bystander less prominent, a lower value keeps the photo looking natural.

Check the result at the size it will be seen. A blur that hides a face on a phone screen may be less convincing on a large monitor.

The result keeps the format of the photo you added, so a JPG comes back as a JPG and a PNG as a PNG, saved at high quality. The blurred pixels replace the originals; the face is not stored underneath and cannot be recovered from the file. The download also carries no EXIF data, so the camera model, the date and any GPS position recorded by a phone are gone as well.

You can add a whole set of pictures. Each one is examined separately and appears in the strip under the preview, where you can open it, review its boxes and adjust them. The effect and its strength apply to the whole set. When every picture has been checked, one button downloads them together as a ZIP file.