Pixelate & Blur Tool

About Pixelate & Blur Tool

This tool hides parts of a photo with pixelation, blur or a solid fill. You can draw any number of regions as rectangles, ellipses or brush strokes, and each region has its own effect and strength. Pixelation divides the region into blocks (4–60 px, default 16) and paints each block with the average color of its pixels. Blur runs a box blur of the chosen radius (2–40 px, default 12) three times, which approximates a Gaussian blur. Solid fill covers the region with an opaque box in the color you choose (black by default).

Editing happens on a copy scaled so its long side is at most 1400 px; the download is recomputed at the original resolution. Region coordinates and strength are scaled by the same factor, so on a 4000 px photo a 16 px block becomes roughly a 46 px block in the saved file. That keeps the result looking as coarse as the preview.

Pixelated text can be recovered

Pixelation does not delete information; it averages it. Pixelate the same character, in the same font and size, at the same block position, and you always get the same colored blocks. Depix, published by Sipke Mellema in 2020, used exactly this to recover pixelated passwords from screenshots of text editors: it pixelates candidate characters under the same conditions and looks for blocks that match.

Blur is not safe either. A 2016 study by McPherson, Shokri and Shmatikov trained neural networks to identify pixelated and blurred faces and digits, and they succeeded far more often than humans or chance. Blur mathematically mixes neighboring pixels, which is especially weak when the hidden content comes from a small set of possibilities, like digits or letters.

The practical rule: for text-based secrets such as passwords, account numbers, ID numbers, addresses or API keys, cover them with an opaque solid rectangle, not pixelation or blur. In this tool, set the effect to Solid fill: every pixel in the region is overwritten with one color, so nothing of the original text survives in the saved file.

Why this tool's pixelation needs extra care

Each time you create a region, this tool shifts the start of its block grid by a random amount within one block. That is harder to reproduce than a grid pinned to the region's top-left corner, but an attacker can still try every offset. Making blocks much larger than the text height, so several characters share one block, makes recovery much harder, but the rough shape of the text, its length and line layout still show through.

Also remember that strength scales with the original size. On a small screenshot, one with a long side of 1400 px or less, the number you set is applied as is, and a 16 px block may cover only one or two characters of body text.

Faces, license plates and what to use for each

For bystanders' faces or a shop sign in the background, where the goal is only to make identification difficult, pixelation or blur is reasonable. Make the blocks large enough that the eyes, nose and mouth fall into one or two blocks; with small blocks, people who know the person will still recognize the outline and hair.

License plates are short strings in a fixed format, which is the ideal case for recovery attacks. If it matters, cover them with a solid box. Blur looks smoother, but at a small radius the original shapes show through; the radius should be at least the character height.

What to check after redacting

Because the tool redraws the image on a canvas and encodes a new file, EXIF metadata such as GPS location is not carried over. The same information often appears elsewhere in the picture, though: a reflection in a window, a shipping label, a monitor in the background. Zoom into the saved file and scan it once more.

Your original file is unchanged. Share only the redacted copy, and be careful not to upload the original alongside it, for example in the same cloud album.

Redaction mistakes that are not about the image

Screenshots of documents are a common source of leaks. If the source is a PDF or a web page, the hidden text may still be selectable or searchable in the original, so redact a screenshot rather than sharing the file with shapes drawn over it. And if the same text appears twice, in a title bar and in the body for example, both copies have to be covered.

Worked examples

How to use it

1. Load an image and pick the rectangle, ellipse or brush tool.
2. Draw over what you want hidden and choose pixelate, blur or solid fill and its strength (or color) for each region.
3. Choose a format and quality and download at full resolution.

Frequently asked questions