Hiding faces
Choose the photos
Pick up to 20 at once. Thumbnails let you move between them.
Find the faces
Choose Fast (Tiny Face Detector), a 1.5 MB first download, or Thorough (SSD MobileNet v1), 6.7 MB, which finds more small or turned faces. Press Find faces, or Find in all for the whole set.
Correct the boxes
Drag on the picture to add a box the detector missed, such as a number plate or a name badge. Click a box to switch it off; it then shows as dashed.
Choose the effect
Blur, Pixelate or Solid block, with a strength slider for the first two. Grow found faces by widens automatic boxes to cover hair and ears, and the shape can be Oval or Rectangle.
Save
Pick JPG, PNG or WebP. Blur this picture saves one; Blur all and download ZIP saves the set.
Example: photos from a networking breakfast
A small agency hosted a breakfast for local business owners and wants to post twelve photos. Two guests asked not to be shown, and a few people in the background did not know they were in shot.
The Thorough detector is chosen because the photos include groups. After Find in all, the boxes of the guests who agreed to appear are clicked off, one face at the edge of a group shot that was missed is added by dragging, and the background faces stay boxed. The ZIP holds every photo with its file name and -blurred on the end.
12 photos, breakfast-01.jpg to breakfast-12.jpg Thorough (SSD MobileNet v1), Blur, strength 12, grow 20%, Oval, JPG
blurred-photos.zip containing breakfast-01-blurred.jpg to breakfast-12-blurred.jpg
Detection is a helper, not a guarantee
Automatic detection misses some faces: people in profile, very small faces at the back, faces partly hidden by a hand or a hat, and anyone in low light. Look through every picture at full size before publishing and add boxes by hand where needed; the count of faces found in each picture is a useful prompt.
A photo in which someone can be recognised can be personal data under UK GDPR, so hiding faces is a sensible step before sharing event, office or site photos publicly; the ICO's guidance explains when that applies. Blurring a face does not hide everything, since clothing, tattoos and context can still identify someone.
Questions people ask
Are my photos uploaded to detect the faces?
No. The open-source face-api library and its model files are downloaded once from the jsDelivr code library when you first press Find faces, and then run inside your browser. Your photos are read and changed on your own device and are never sent anywhere.
Is pixelating safer than blurring?
Strong blur and coarse pixelation are both hard to reverse in practice, but light versions of either can leave a face recognisable. Solid block removes the information completely and is the safest choice for anything sensitive.
Can I blur number plates or name badges too?
Yes. The detector only looks for faces, but you can drag a box over anything else, such as a plate, a badge or a screen. Boxes you draw yourself are hidden exactly as drawn and are not enlarged by the grow setting.
Where your data goes
Photos are processed in your browser and are not uploaded. On first use the page fetches the open-source face-api code and model weights from cdn.jsdelivr.net; nothing about your pictures is sent back.
Checked against the tool on