Blur a Face, Text and Personal Data in a Photo

Load the picture — faces are found and covered as the file opens. If you specifically want blur, pick the Blur mode: each box can be moved, resized and deleted. For faces blur is usually enough; for documents and text prefer fill, because blurred text can sometimes be recovered.

Blur averages the pixels of an area but leaves a trace of the original in them, and that trace is what lets text be read back. A solid fill simply overwrites the pixels: the original data is not in the new file at all. Pixelation here sits in between — it shuffles the pixels before building the mosaic, so the blocks carry no recoverable signal either.

It depends on how it was covered. Blurred and plainly pixelated text is sometimes recovered: the Depix and Unredacter projects demonstrated it on real screenshots by searching candidate renderings and comparing the block pattern. Text under a solid fill cannot be read by any method — the original pixels are simply not in the file. That is why fill is the recommendation for passports, cards and chats.

Load the screenshot — identity, tax and card numbers are found automatically. Numbers are validated by their own checksums, so prices, dates and order numbers are not touched. Review the boxes in the list, add your own where needed, and save: fill overwrites the pixels.

Card, tax and registration numbers carry a checksum — the last digits are computed from the earlier ones. The tool runs that arithmetic: a real number passes, a random one (a price, an order id, a part number) does not. Distinctive shapes such as a 4-4-4-4 card number or a licence plate are also recognised by appearance, so nothing is missed because of an OCR slip.

Plates are recognised by their shape — a fixed pattern of letters from a restricted alphabet plus digits and a region code. A plate lands in the found list alongside faces and documents, and you can simply leave it covered. If the plate is at a sharp angle or motion-blurred, draw a rectangle over it by hand.

The tool builds a new file from the pixels on the canvas, so GPS coordinates, capture date, camera model and the embedded thumbnail do not reach the result. Before downloading you can see which fields were present in the original — the case where a carefully covered picture could still have said where it was taken.

Yes: select several files or add them one by one and a queue appears. “All files” runs the automatic search across the whole batch, and the result comes back as a single ZIP where every image keeps its own resolution and format.

The format and resolution of the input: JPG → JPG, PNG → PNG, WebP → WebP. HEIC photos are saved as JPG. You can change the format by hand next to the download button.