Pixelate Image

Aim for the face itself spanning no more than 6 to 8 blocks across. On a photo where a face is 400 pixels wide, that means a block size around 50 to 64. If the face still reads as a face when you squint, the blocks are too small.

Each block averages detail away permanently, so exact recovery is impossible. But with tiny blocks, attackers have matched mosaics against candidate faces or known text by pixelating guesses and comparing, and machine learning models can reconstruct convincing approximations. Bigger blocks destroy more information and make all of that far harder.

The tool pixelates the whole frame. To mosaic one region only, crop the sensitive area with the Image Cropper, pixelate the crop here, and paste it back over the original in any editor. It takes a minute and keeps the rest of the photo sharp.

Pixelation is generally the safer censor: a block of averaged colour carries less recoverable structure than a smooth blur of equal coverage, and it visibly signals that something was redacted. Blur looks nicer for backgrounds and decorative effects. For plates and documents, pick pixelation with a generous block size.

The canvas re-encode this tool uses doesn't copy EXIF into the output, so GPS and camera tags get dropped as a side effect. Still, run the downloaded file through the Image EXIF Viewer before posting anything sensitive, so you're confirming rather than assuming.

Human vision is very good at faces, and a mosaic preserves the overall colour layout of skin, hair, and clothing. Increase the block size until identifying features vanish, and check the result at thumbnail size too, since shrinking an image is exactly how people defeat weak pixelation.

Yes. Add multiple files and bulk mode processes them sequentially with the same block size, returning pixelated-images.zip. That's handy when a set of app screenshots repeats the same sensitive region in the same spot.