Pixelate an image or part of one
Pixelation replaces each block of pixels with one flat colour, destroying the detail rather than smearing it. Use the region controls to cover a single face or a name badge and leave the rest of the photo untouched. Everything is processed on your device.
How to use the Pixelate Image
- Add the photos you need to anonymise.
- Set the block size — larger blocks hide more detail.
- Leave the region at 0/0/100/100 to pixelate everything, or set a rectangle in percentages to cover just one area.
- Click Pixelate, check the preview, and download.
How pixelation works here
The selected rectangle is scaled down so that each block of pixels becomes a single pixel, then scaled straight back up with smoothing switched off. Each output block therefore carries one colour, averaged from everything that was inside it. That averaged value replaces the originals in the saved file: the fine detail is gone from the output, not merely covered by an overlay that could be peeled off. This is the important difference between pixelating an image and drawing a shape on top of it in a layered editor, where the original layer often survives in the saved document.
Choosing a region in percentages
The four region fields describe a rectangle as percentages of the image, which keeps the same settings usable across a batch of different sizes. Left 60, top 10, width 25, height 30 covers a face in the upper-right quarter of any photo with a similar composition. Percentages are clamped to the image, so a region that runs off the edge is trimmed rather than failing.
How big should the blocks be?
| Goal | Guidance |
|---|---|
| Hide a face | Block size such that the face is 4–6 blocks wide. On a 1500px photo with a 200px face, that is around 35–50. |
| Hide a car number plate | At least one block per two characters — plate glyphs are small and regular, so err large. |
| Stylistic 8-bit effect | 8–24 on a normal photo, with the region left at full frame. |
| Hide printed text | Do not. Use a solid block instead — see below. |
The limits of pixelation for text
Pixelating a photograph of a face at a sensible block size is genuinely hard to undo, because faces vary continuously and there is no dictionary to search. Text is different. If an attacker knows the font, the size and the character set, they can render every candidate string, pixelate it the same way and compare — and published tools have recovered pixelated text this way. Treat pixelation as safe for faces and scenes, and never as a redaction method for names, numbers, addresses or codes. For those, crop the area out or paint over it with a solid opaque rectangle before you export.
Also remember the metadata
Hiding a face does not hide where the photo was taken. A JPG straight from a phone usually carries GPS coordinates and a timestamp. The output of this tool is redrawn from pixels, so that metadata is not carried through — but if you publish the original alongside it, the location is still there.
Frequently asked questions
Can pixelation be reversed?
Not reliably for faces and scenes. For text it can be, if the font and layout are predictable — several published tools do exactly that. Use a solid block for words and numbers.
How do I pixelate only a face?
Set the region fields to the rectangle around the face, as percentages of the image width and height, and raise the block size until the features are unrecognisable.
Is the original detail still inside the file?
No. The averaged blocks replace the original pixels before the file is written, so there is nothing underneath to recover.
Can I pixelate many photos with the same settings?
Yes. Because the region is given in percentages, one setting works across a batch of differently sized images with similar framing.
Privacy
This tool runs entirely inside your browser using WebAssembly and the Canvas/File APIs. Your files are never uploaded to ToolFlint or any third party; you can verify this in your browser's network tab or by switching to airplane mode after the page loads. Read how we process files.
Last updated 2026-09-23.