Pixelate faces
Covers faces with a mosaic instead of a blur, so it is obvious the photo was edited on purpose. Runs entirely in this browser tab — the photo is never uploaded.
Face pixelate editor
Sample photos
0 mosaic areas. Tap an area to adjust it.
Mosaic or blur — which one to use
A mosaic cuts the area into coarse squares and flattens each square to a single colour. Blur the same area instead and the result can read as a photo that simply missed focus, which leaves the viewer unsure whether you covered something deliberately or your hand shook. When the point is that the covering is visible, a mosaic says so. Marketplace listings, photos attached to a complaint or a claim, and material circulated inside a company are all cases where you want that to be unambiguous.
When you would rather keep the photo looking like the photo, use face blur instead. Either way, how much survives depends on the area you choose and the strength you set.
What the pixel size number means
The number is not the width of a square in pixels. It sets how many squares the short edge of the selected area is divided into. That is why a small face and a large face end up about equally unreadable rather than the small one turning to mush while the large one stays legible.
| Pixel size | Squares across the short edge | On an area 200px across |
|---|---|---|
| 10 (finest) | about 23 | each square about 9px |
| 48 (default) | about 9 | each square about 24px |
| 85 (coarsest) | about 5 | each square about 38px |
Nine squares across a face removes most of the detail, but that is not a guarantee nobody can tell who it is. Push the setting fine enough — past roughly 20 squares — and the outline of the face comes back. Finer means closer to the original, which is the opposite of what you want when you are covering something. If you are hiding rather than decorating, go coarse.
Where people get caught out
Pixelated text cannot be read
Text is the risky case. Name badges, addresses and account numbers can keep enough of their outline through a mosaic to be guessed. Letters are a handful of thick strokes, so even a moderately fine grid leaves the shape recognisable. For text, push the pixel size to the maximum, or use the solid fill in hide license plates, which covers completely.
Covering the faces is enough
Detection only looks for faces. ID badges and lanyards, documents on the desk, a screen that is switched on, a nameplate on a door, an address label on a parcel — all of that goes out exactly as photographed. Tap an empty spot to add an area anywhere, not just over a face.
Detection finds every face
Profiles are missed by design. The detector is built for front-facing faces at a reasonable size. A head turned well to the side, a face only a few pixels across in a large group, or a face half cut off at the edge of the frame will be skipped. Count the people in the photo before you save.
The file you download contains the pixelated image, and it does not carry the original photo's metadata — no capture location, no camera or phone model. Your original file stays on your device untouched, so check you are attaching the one with -face-pixelate at the end of its name.