Free photo tools/Cover or pixelate part of a photo

Is pixelation safe for hiding text?

No. Pixelation and blur are not safe for text, numbers, or license plates. Both compute the new pixels from the old ones, so the result still depends on what was written there, and software can work back from it. A solid cover replaces the pixels and leaves nothing to work back from.

By Paul Bădărău, maker of Milkglass. Written on 7 October 2026.

What pixelation does

Pixelation, also called a mosaic, lays a grid over an area. Each square of the grid gets one color: the average of all the pixels that were in it. A line of text becomes a row of gray squares, some lighter and some darker.

To your eye the text is gone. In the numbers, part of it remains. A square that covered the thick stem of a letter is darker than a square that covered a gap. The pattern of light and dark squares is a coarse fingerprint of the text.

How text is recovered from it

Nobody "sharpens" the squares back into letters. The method is to guess and compare:

1. The attacker finds out the font and the size. In a screenshot that is easy, because the text around the hidden part uses the same ones.
2. A program types one candidate character, pixelates it with the same grid, and compares the squares with the squares in your picture.
3. The best match stays, and the program goes on to the next character.

This works because pixelation is plain arithmetic: the same text always gives the same squares, and a change in one letter changes only the squares near it. Dan Petro of Bishop Fox, who published a tool for it in 2022, describes exactly this property and draws the conclusion: "When you need to redact text, use black bars covering the whole text. Never use anything else." (Bishop Fox, about Unredacter).

What the research found

The warning is older than that tool. In 2016, Steven Hill, Zhimin Zhou, Lawrence Saul, and Hovav Shacham studied the two most popular ways to hide text in screenshots, mosaicing and blurring. They used statistical models of the kind that speech recognition uses and recovered "both short and indefinitely long instances of redacted text". Their conclusion: mosaicing and blurring "are not viable approaches for text redaction" (Proceedings on Privacy Enhancing Technologies, 2016, issue 4).

Is blur any better?

No. Blur replaces each pixel with a weighted mix of its neighbors. It is arithmetic too, and the 2016 study covers it together with the mosaic. A strong blur over large text hides more than a light one, and you cannot judge by eye how much is enough. That is the real problem with both methods: they look safe long before they are.

What is safe: a solid cover

A solid cover gives every pixel in the box the same color, whatever was under it. Two different names under two boxes of the same size give exactly the same pixels. With nothing of the text left in the picture, a program has nothing to compare its guesses with.

Three conditions make a solid cover hold:

• It is fully opaque. A see-through marker darkens the text and keeps it.
• It is in the pixels. The file you send is one flat picture, with no layer under the box.
• It is large enough. It covers the whole text with a margin, and it hides how long the text is when the length is a hint.

The photo redactor on this site works this way. Solid cover is its default. It paints each box black into a new picture and saves that picture as a new file.

What pixelation is fine for

Pixelation is a good tool for a different job: to make a part of a picture quiet when nobody has a reason to work on it.

Fine

A passer-by in a holiday photo. A messy background. A screen or a poster behind you. A logo. Anything where the picture should stay pleasant and the hidden part is of no value to others.

Use a solid cover

Names, addresses, phone and account numbers, passwords and codes, license plates, barcodes and QR codes, signatures, and the face of a person who must stay unknown.

The rule of thumb: if the hidden part is made of characters, or if someone could want it enough to try, cover it.

How coarse is the pixelate of this tool?

The redactor has a pixelate option for the "fine" cases above. It is coarse on purpose: at most four squares across the short side of a box, and no square smaller than 16 pixels of the photo. A box around one line of text gets at most four rows of large squares. The tool still names the solid cover as the safe choice for text, and so do I.

If you already posted a pixelated screenshot

Treat the hidden text as known to anyone who cares enough. Replace the picture with a covered one where you can, and change what can be changed: a password, a code, a link that gives access. For the next one, redact the screenshot with a solid box.