How to blur faces in a photo before you post it

To blur faces in a photo before posting it, run the picture through a face finder, check every box it draws, and choose an effect that hides each face completely. Solid is the safest effect, because it replaces the pixels inside each box with one colour. The tool below does this in your browser, so the photo is not uploaded, and the download is redrawn without camera and location data.

Blur facesFaces need hiding in a photoOpen

When blurring faces is the right call

People appear in many photos they did not choose to be in: a bystander in a street shot, a child at a school event, a person at a protest, a customer in a shop. A clear face makes that person recognisable to anyone who knows them, and for children and for people at a demonstration, being identified can have real consequences. Blurring the face keeps the scene readable and removes the one thing that identifies the person.

Rules on publishing identifiable photos of people differ by country and by setting. Schools, clubs and workplaces often have their own rules about photos of children or staff, so check those before you post. A sensible default is to hide the face of anyone who has not agreed to appear.

How detection works, and where it misses

The tool uses an on-device face detector, the Ultra-Light-Fast-Generic-Face-Detector-1MB in its RFB-640 version, which is released under the MIT licence. It draws a box around each face it finds. It also checks a mirrored copy of the photo, which gives faces turned one way a second chance. Very large photos are checked again in overlapping quarters, because a small face in a big picture can be too small to find at the detector's input size.

Detection is not a guarantee. The detector finds most faces turned roughly towards the camera. Very small, turned, partly covered or blurry faces can be missed, and so can small faces in the background of a crowd. Sensitivity has two settings. Normal finds the clear faces. High looks harder, catches more borderline faces, and can add a few boxes that are not faces. You can leave any box off the blur by tapping it, and if the detector missed a face, drag across the photo to cover it with a new box. Undo box removes the most recent box you drew by hand.

Blur, pixelate or solid

The effect decides how well the face is hidden.

• Blur averages the pixels in the box into a soft smear. A strong blur hides a face from people looking at the photo.
• Pixelate replaces the area with large blocks of averaged colour. It also hides a face from viewers, and the block pattern is easy to see.
• Solid covers the box with one flat colour. The pixels inside it are replaced, so no facial detail is left in the file.

For an ordinary bystander in a holiday photo, a strong blur or pixelation usually hides the person well enough. Both still keep traces of the original pixels, though, and fine details can sometimes be partly reconstructed. For a face you never want anyone to see again, choose Solid. The strength slider controls how heavily Blur and Pixelate work, so set it high enough that you cannot pick out features in the preview.

The shape only changes the outline of the cover. Oval fits the outline of a head. Box covers a plain rectangle, which can be easier to size over a face that is turned or partly hidden.

Location is a separate problem

A phone photo can record the place it was taken, the camera model and the date. The download from this tool is a new image drawn from the pixels, so those details from the original file are not carried over. Blurring the faces does not remove them from the original, and you should not share the original.

Location can also appear in the picture itself: a street sign, a house number, a school name or a shop front. A blur does not hide those, so read the whole picture before you post it. If you are working with a photo from another tool, or want to see what a file contains, remove the location data separately.

Other details that identify people

Faces are not the only identifiers. A name tag, a licence plate, a tattoo, a logo on a jumper or a phone screen can each identify someone. For those, use solid boxes. Redact the image draws solid boxes over names, numbers or addresses, and the same rule applies: a box on a flattened image works, while a semi-transparent highlight does not. The screenshot guide explains why.

Clothing, body shape and the setting can also identify people, especially when a caption names the place or the event. Hiding a face is one step, not full anonymity.

Check the result before you post

1. Look at each box. Every face you meant to hide should be covered, and every face you left visible should be one you chose to show.
2. Zoom to 100 percent on the edges of each blur. A box that is slightly too tight can leave the chin or hairline showing.
3. Look for reflections. A face in a window, a mirror, a car door or a phone screen can be just as identifying as the face itself.
4. View the picture at the size and in the app where people will see it. Feeds often resize or crop images.
5. Check the file name and any caption. A name in either one undoes the blur.

Quick answers

Does the photo get uploaded? No. The face model is downloaded to your browser once, and the photo is processed on your device.

Will it find every face? No. It finds most faces turned towards the camera and can miss small or partly hidden ones. Check every photo, and drag a box over any face it missed.

Is blur safe enough? For ordinary privacy, a strong blur or pixelation hides a face from viewers. For anything sensitive, choose Solid.

Does the download keep the GPS location? No. The download is redrawn from the pixels, so camera details and GPS coordinates from the original file are not carried over.