Pixelate Image
Pixelate an image or video into square blocks, fine to huge, across the whole frame or only inside a box or ellipse you place. Free, and nothing is uploaded.
What does it mean to pixelate an image?
Pixelating an image means showing it at a much lower resolution, so each area becomes one square block of a single color. TV news uses it to blur faces. Westworld (1973) was the first feature film to pixelate photography digitally, to show an android's point of view.
This image pixelator turns a whole photo or video into even square blocks, anywhere from barely visible to huge.
How to pixelate a photo or video
Load an image or a video and set Size, the block width, from 1 to 100 pixels of a 1080p frame. It's counted down the frame: Size 10 gives about 108 rows of blocks on any picture, and a wider frame just gets more columns.
Colour from decides what each block shows. Centre takes the block's middle pixel and keeps colors crisp. Average blends the whole block for softer tones, and it flickers less on video, since a block only changes as much as its whole area does.
The presets jump between the common looks: Blocks, Fine, Chunky, Huge and Smooth average. If you also want the blocks snapped to a retro palette, open Photo to Pixel Art; for a honeycomb instead of squares, try Hexagon Mosaic.
Settings for a hidden face, a blocky thumbnail and a sharp subject on a pixelated background
Hiding a face or a number plate: Huge sets Size 64, which is about 17 blocks down the whole frame, and Colour from Average keeps one bright pixel from standing out in a block. Then set the effect's Mask to Ellipse or Rectangle and drag it over the face, so the rest of the picture stays sharp.
Blocky thumbnail: Chunky (Size 28) leaves about 39 rows. That's coarse enough to make small text unreadable while big shapes stay clear, and Centre keeps the colors bright.
Sharp subject, pixelated background: pick Ellipse under Mask, cover the subject with it and tick Invert. Now only the area outside the ellipse breaks into blocks, and Feather softens the edge between the two.
Can pixelated text be read again?
Often, yes. Bishop Fox recovered text hidden under 5-pixel blocks with a tool that guesses one character at a time, pixelates each guess the same way and keeps the closest match. A 5-pixel grid can only line up 25 ways, so the search took minutes. Depix, an open-source tool, matches blocks against a screenshot of known text in the same font, and Hill and colleagues read text back even when the blocks were bigger than the letters.
Colour from Average blends 16 evenly spaced samples from each block. That's close to the plain block average Depix was written for, and every bit as predictable. Size scales with the picture's height, so on a screenshot 1,080 pixels tall, Size 5 gives exactly the 5-pixel blocks Bishop Fox read back; the same setting on a taller phone screenshot gives bigger ones.
For a password, an address or a name, cover the text with a solid box in an image editor. The censor tool on the Image Tools hub pixelates as well, so it has the same limits.
How big do the blocks have to be to hide a face?
Bigger than most people go, and even then it can fail. McPherson, Shokri and Shmatikov trained a simple neural network on pixelated faces. With face photos 14 blocks across, it picked the right one of more than 530 celebrities 58 percent of the time, against 0.19 percent by chance. Faces only 6 by 7 blocks, from a set of 40 people, it named 96 percent of the time.
Count the blocks down the face in your own picture: its height in pixels divided by the block size, and the block size is Size times the picture's height divided by 1,080. A face 300 pixels tall in a 1080p frame is 30 blocks tall at the default Size 10, about 11 at Chunky and under 5 at Huge. Start below the study's 6 by 7. And if a viewer could know the person, a solid patch is the only cover you can count on.
Video is weaker still. The blocks shift a little every frame as the face moves under them, and Cavedon and colleagues rebuilt pixelated video by combining frames. The Mask also stays where you put it, because its box can't be keyframed, so make it big enough to cover the face's whole path.
Saving a pixelated picture
The saved file holds only the rendered pixels. A still taken from the page's canvas carries none of the original's EXIF data, so camera details and GPS location don't come along, and nothing from under the blocks gets stored.
PNG keeps block edges hard. JPG adds faint ringing along them, which doesn't matter on a hidden face but shows on large flat blocks. A pixelated clip saves as MP4 or WebM, where Average flickers less and gives the encoder less to store.