Blur faces in photos
Make people in photos unrecognisable before you share them: the face detection finds the faces automatically – even small ones in the background of group photos. Pixelate, blur, a bar or an emoji; other spots such as number plates or name tags you simply drag across yourself. Everything happens in your browser, with no upload and no third parties.
Face detection YuNet · one-time ≈ 17 MB from mytoolio.org
✓ Metadata such as GPS location and camera data is removed when saving.
No photo yet. Drop an image here – faces are found and pixelated automatically.
Tap a frame to switch it off or on. Drag across the image to add an area – e.g. for number plates or name tags. Frames can be moved; the bottom-right corner resizes them.
How it works
Choose photos or a video
Drop one or more photos – or a video – into the box, pick them or paste them with Ctrl+V. They are only read locally.
Check the faces
Faces are found and covered automatically. Tap a frame to leave a face visible, or drag across the image to add your own areas – e.g. for number plates.
Adjust & save
Choose effect, strength and shape and save the image – several photos at once as a ZIP. Metadata such as the GPS location is removed in the process.
Frequently asked questions
Are my photos uploaded?+
No. Your photos are only read in your browser, searched for faces and edited there – they never leave your device. The face detection (YuNet) and its runtime come from mytoolio.org itself; there are no requests to other services.
Does it really find every face?+
Usually, yes – including small faces in group photos and crowds. Hard cases are faces in profile, from behind, heavily covered or very blurry ones. So check every image: a higher sensitivity finds more, and you cover anything missing with a frame of your own.
Which effect is the safest?+
A bar is the safest because it covers everything. Pixelate and blur are built here so that only a few coarse patches of colour remain – so even the blur cannot be reversed. The higher the strength, the less stays recognisable. Keep in mind: clothes, hair, tattoos or the surroundings can still give a person away.
Can I cover number plates or other things too?+
Yes. With “Also find text & number plates” the tool finds clearly readable plates, signs and text automatically (it misses very small ones). For anything else: just drag a frame around the spot in the image – for example a number plate, a name tag, a house number or a screen. Frames can be moved and resized from the bottom-right corner; delete your own frames with the Delete key or tap them to switch them off.
What happens to the metadata?+
When saving, the image is created anew. All metadata is dropped in the process – including the GPS location, camera model and time taken. The original photo on your device stays unchanged.
When should I blur faces?+
In many countries, publishing a photo in which people are clearly recognisable needs their consent (in the EU, for example, under the GDPR). Crowds at public events are often an exception. Blurring is the simplest way to be on the safe side. This is not legal advice.
Can I blur videos too?+
Yes – drop a video (MP4, WebM, MOV) into the box and press “Blur video”. First all faces are found and followed over time, then the video is written anew. With MP4/MOV this happens frame by frame faster than the video plays (WebCodecs): the sound is taken over unchanged, and the new file holds no original pictures and no metadata such as place or device. Other formats are recorded again in real time (about twice as long as the video; keep the tab in the foreground). Afterwards “Check faces” lets you go through the video: tapping a face switches it off or on for the whole video, a dragged rectangle covers a spot for the whole video – then “record again”. Check the result before sharing, especially with fast movement and faces in profile.
Face detection with YuNet (OpenCV Model Zoo, MIT licence) via ONNX Runtime Web (MIT) – both served by mytoolio.org and run entirely in your browser. The detection can miss faces; please check every image before sharing.