Skip to content

Privacy Blur

Detect faces in a photograph automatically and pixelate them before you share it.

Privacy Blur

Automatically detect and blur faces in photos.

What Privacy Blur does

Publishing a photograph taken in a public place regularly means publishing the faces of people who did not agree to it. Obscuring them is the straightforward answer.

A face-detection model runs in the page, finds the faces it can, and pixelates each one. The model files are served from this site rather than from a third party, so no external service sees your photograph.

  • Automatic face detection using a lightweight model
  • Pixelation applied to each detected face
  • Exports as PNG

How to use Privacy Blur

  1. 1

    Add your photograph

    The detection model loads on first use.

  2. 2

    Run the detection

    Faces that are found are pixelated in place.

  3. 3

    Check every face

    This is the important step. Confirm that each face you wanted obscured actually has been before you share the image.

  4. 4

    Download the result

    The pixelated version is exported as a PNG.

Limits and known behaviour

  • Detection is not reliable enough to trust blindly. Faces in profile, partially covered, small in frame, at unusual angles, or in poor light are regularly missed. Always check the result yourself.
  • Only faces are detected. Name badges, licence plates, house numbers, tattoos and reflections are left untouched, and any of them can identify someone.
  • Pixelation strength is fixed and cannot be adjusted.
  • There is no manual selection, so a face the model misses cannot be blurred here.
  • The pixelated image is a new file; the original on your disk still contains the faces.

Privacy and data handling

Runs entirely in your browser

  • Your image is decoded and redrawn by the browser inside this page. It is never uploaded.
  • The face-detection model is served from this site, so your photograph is never sent to an external recognition service.
  • Detection identifies where faces are. It does not recognise who anyone is, and no biometric data is produced or kept.
  • Nothing is stored; the file and the result disappear when you close the tab.

Site-wide data handling, including analytics and advertising, is described in the privacy policy.

Frequently asked questions

Can pixelation be reversed?

Not from the exported image, since the original pixel values are gone. It is worth knowing that weak blurring of small, predictable content — such as a short number — can sometimes be defeated by other means, which is why heavy pixelation of a face is a reasonable approach and a light blur is not.

Why did it miss a face?

Lightweight detection models are tuned for reasonably frontal, reasonably sized faces in decent light. Profiles, distant figures, sunglasses, masks and motion blur all reduce detection. Check every image before publishing.

Does this identify the people in my photograph?

No. It locates faces; it does not match them against anyone. No identity information is produced.