Pixelate / blur region

Not from the pixels that were replaced. The region is overwritten in the pixel buffer and the whole image is encoded from scratch, so there is no hidden layer and no original underneath, unlike a black rectangle drawn on top of a PDF. What you keep on your own device is a different matter: your input file is untouched and still holds the original.

No. There is no detection of any kind. You give the rectangle yourself as four numbers: Left edge and Top edge of the region, both default 0, plus Region width and Region height, both blank by default. Leaving the size blank stops with "give both a width and a height for the region".

Redaction style defaults to pixelate, which averages each block of Mosaic block size pixels, default 10, into one flat colour. Blur mode instead box-blurs each pixel over Box-blur radius, default 4. Pixelate with a large block destroys the most detail. A small radius or a small block leaves recognisable structure, so increase them for anything sensitive.

Because the rectangle runs past the edge of the picture. The left edge plus the width must fit the image width, and the top edge plus the height must fit the image height, and the message repeats your numbers and the real image size. All four values must be whole numbers of 0 or more.

One rectangle per run, so run the tool again on the output for a second area. The rest of the image is left exactly as it was, except that the whole file is re-encoded, at the Quality option, default 80, for JPG and WebP. The result is named with a -redacted suffix in the original format.

No. The image is decoded, edited and encoded in a Web Worker in this tab with the jSquash codecs. Nothing is uploaded, there is no account, and the page keeps working offline after the first visit, which matters when the thing you are hiding is the reason you came.