Blur Faces in Photo

Before publishing a photo of a crowd, a classroom, a street or a workplace, you often need to hide the faces of people who did not agree to be shown. Privacy laws, school policies and newsroom ethics frequently require it. Blurring dozens of faces by hand is tedious; this tool finds them automatically on your device and lets you correct the result before saving.

How faces are found

The page uses YuNet, a compact face-detection model of about 230 kilobytes, running in your browser. It first scans the whole picture to find large faces, then examines overlapping tiles of the photo at higher resolution to catch small faces in the background. In large group photos, faces down to roughly 12 pixels across can be detected. The photo is never uploaded.

Choosing the effect

Blur softens faces into an unrecognizable smudge while keeping the photo pleasant to look at, the usual choice for social media and news. Pixelate replaces faces with coarse blocks, a familiar signal that something was hidden on purpose. Black box covers faces completely and is the safest option when identification must be impossible. Strength controls how intense blur and pixelation are.

Shape and margin

Oval shapes follow the outline of a head naturally; rectangles are simpler and cover more. Extra margin enlarges each area around the detected face, which helps cover ears, hairlines and chins that detectors box tightly. A margin of 20 to 30 percent is a sensible default for anonymization.

Sensitivity

Higher sensitivity finds more faces, including small or partly turned ones, at the cost of occasional false areas on patterns that resemble faces. Above 75, the slider warns that false detections become likely. False areas are easy to remove: click them to switch them off.

Correcting by hand

Automatic detection is a starting point. Drag on the photo to add an area over any face that was missed, or over anything else that should be hidden, such as a license plate or a name badge. Click an existing area to switch it off. Always look through the whole picture before saving.

What gets saved

The download is a new image containing the blurred areas burned into the pixels; there is no layer or undo hidden in the file. Location, camera details and other metadata of the original are not carried over.

Limits to keep in mind

Detection can miss tiny, side-on, masked or very dark faces. People can also be identified by clothing, tattoos, voice in videos, or context such as a house number. Anonymizing well means thinking beyond faces. For very sensitive material, combine black boxes with cropping.

Typical uses

School newsletters, event recaps, street photography, research datasets, screenshots of video calls, real-estate photos with people in them, and citizen journalism.

Consent and the law

In many countries, recognizable people in published photos have rights over their image, and data protection rules such as the GDPR treat faces as personal data. Schools and employers often need written consent before showing someone. Blurring is the practical alternative when consent is missing. When in doubt, hide the face; a blurred bystander costs a story nothing.

Related tools

To blur any region freely, use Blur Image. To remove hidden metadata from photos you do not need to edit, the EXIF Data Remover cleans them without re-encoding. For documents, Redact PDF removes text permanently.