Blur faces in a group photo

A face detector works on patterns of light and shadow that eyes, a nose and a mouth produce. In the front rows that pattern is clear. Towards the back, faces shrink until they are a few pixels across, and the pattern dissolves into noise. This tool examines large pictures in overlapping sections to reach smaller faces, but there is always a size below which a face is not found by any detector.

Faces turned sideways or tilted far back, faces under hats that shade the eyes, and faces partly outside the frame are the other usual misses.

A hidden face does not always make a person anonymous. In a small group, people who know each other can identify someone from height, clothing, hair or the place they are standing. A school class with one face left clear and the others blurred is still recognisable to every parent of that class. Consider that before posting, and remember the caption and the location tag say a lot too.

Add all the pictures from an event together. Each is scanned as it arrives, and the strip below the preview shows which ones are finished. Open them one at a time, clear the boxes of the people who agreed to be shown, add areas where a face was overlooked, and move on. The blur strength is shared across the set, so the album looks consistent.

When you have been through every picture, download them all as one ZIP. The files keep their names with "blurred" added, so they sort next to the originals without overwriting them.

Each result is saved in the format of its original at high quality, which keeps a JPG album a JPG album. Nothing outside the blurred ovals changes, although a JPG is encoded again when it is saved, as it is by any editor. If file size matters afterwards, run the finished pictures through the compressor.