Photo Censor

Photo Censor finds faces in a photo automatically and blurs or pixelates them — no manual selecting or masking. Detection and redaction both happen on-device with a small machine-learning model; your photo is never uploaded anywhere.

How to use it

1. Drop or browse for a photo.
2. Pick an effect — Pixelate for a blocky, obviously-redacted look, or Blur for a softer effect.
3. Adjust the strength to control how heavy the effect is.
4. If it misses an angled or partially-visible face, raise detection sensitivity and press Censor faces again; lower it if it's flagging things that aren't faces.
5. Press Censor faces, then download the result.

Why on-device detection?

Face detection runs with a small (~230KB) on-device model, entirely in your browser — the photo itself is never sent anywhere. Re-running with a different strength or effect after detection is instant, since only the redraw changes; only adjusting the detection sensitivity re-scans the photo.

Pixelate or Blur — which should I pick?

Pixelate makes it unambiguous at a glance that a face has been redacted — the usual choice for screenshots, social posts, or anything meant for a general audience. Blur reads a little softer and less "censored," which some people prefer for casual photos.

What does detection sensitivity actually change?

It's the confidence threshold the model uses to decide something is a face. Raise it to catch more faces — angled profiles, partially obscured ones, faces further from the camera — at the cost of occasionally flagging something that isn't a face. Lower it to keep only the model's most confident matches.

Does it need an internet connection?

Only briefly, to load the small on-device detection model the first time you use the tool. After that, detection and redaction both run locally — nothing about the photo itself leaves your browser.