# Pixelate an image

> Pixelation replaces each square block of pixels with one average color, so detail disappears to the eye but the rough pattern survives. Researchers identified 57% of faces hidden with 16 × 16 blocks, and public tools rebuild pixelated text, so use a mosaic for looks and a solid block for anything that must stay unreadable.

## What pixelating actually does

A pixelate filter splits the area you select into a grid of squares. Every square is filled with the average color of the pixels it covers. Large squares remove more detail, small squares keep more. The result is the familiar mosaic look you see on TV.

That averaging is the weak spot. Each square still carries real information about what was underneath: its brightness, its color, where dark strokes sat. When the hidden thing comes from a small set of possibilities (the letters of a font, a few dozen faces from a known group), a program can try each candidate, pixelate it the same way, and see which one produces the same squares. That is exactly how the public tools that undo pixelation work.

So this page treats pixelation as what it is: a visual effect. It is good at making something less prominent. It is not good at making something unrecoverable.

## When a mosaic is a reasonable choice

Pixelate works well when nobody has a reason to work hard at reversing it, or when there is nothing secret underneath:

- softening a cluttered background so the subject stands out;
- toning down a brand logo or a T-shirt slogan in a casual post;
- giving a meme or a thumbnail the retro "censored" look on purpose;
- obscuring a stranger who wandered into the edge of a product shot, when you simply prefer they are less visible.

It is the wrong choice for anything a determined person would want to read or recognize: names, email addresses, account or card numbers, addresses, passwords, faces of people who need to stay unidentified, children's faces, license plates. For those, the default **Solid block** mode paints the area opaque black. Nothing of the original survives inside a solid block, so there is nothing to reconstruct.

## How to pixelate part of a picture

1. Choose a picture with the file button, or drop it onto the box at the top of the page. JPEG, PNG and WebP open in every current browser.
2. Drag across the area you want to hide. Draw as many boxes as you need, and drag a box to move it or remove one you drew by mistake.
3. Under **Cover with**, pick **Pixelate**. The label beside it reminds you that this mode is decorative and can be undone.
4. Look over the preview at full size. The Cover with setting applies to every box in the picture, so if any box covers text or a number, switch back to **Solid block**.
5. Press **Export**, then use the download link (or the share button on a phone) to save the new copy.

To pixelate a whole picture, draw one box from corner to corner.

## Pixelate, blur or solid: a quick comparison

| Mode | What fills the box | Can it be reversed? | Good for |
|---|---|---|---|
| Solid block (default) | Opaque black | No: the pixels underneath are replaced | Text, numbers, faces that must stay unknown, QR codes |
| Pixelate | Average color of each square | Sometimes, with matching tools and a known font or a known set of faces | Backgrounds, logos, a deliberate look |
| Blur | A soft smear of nearby pixels | Sometimes, for the same reasons | Gentle de-emphasis where nothing is secret |

If you are not sure which row your picture belongs in, it belongs in the first one.

## What the exported file contains

Export draws your picture and the boxes into a brand-new image and encodes it from scratch. Because the file is rebuilt rather than edited in place:

- the covered area holds only the mosaic (or the solid color), not the original pixels;
- the camera's EXIF block, GPS coordinates, XMP and IPTC fields, maker notes, embedded thumbnails and Content Credentials are all left behind, because none of them are copied into the new file;
- a JPEG stays a JPEG, a PNG stays a PNG and a WebP stays a WebP; an HEIC photo opened in Safari comes out as a JPEG;
- the image keeps its full resolution. The editor shows you a reduced preview (scaled so neither side exceeds about 2,048 pixels) so it stays responsive, then maps your boxes back onto the original pixels when it exports.

If your browser cannot produce a file that large, which happens with some very big photos on phones, the export tries again with the longer side set to 4,096 pixels and tells you it did so.

Want the location gone but no boxes at all? That is a different job: [remove location data from a photo](/remove-location-from-photo/) keeps every pixel untouched and only strips the hidden fields.

## Formats, devices and limits

- **Opens:** JPEG, PNG and WebP in any current major browser, on a computer or a phone.
- **HEIC from an iPhone:** only browsers that can display HEIC by themselves (Safari) can open it here. In other browsers the page asks you to convert it to JPG first.
- **Several pictures:** you can pick more than one, and each is edited and exported on its own.
- **Animated GIFs and video:** not supported. Only still images.
- **Block size:** set automatically, at about 12 blocks across the shorter side of each box and never under 4 pixels, so a small box and a large box both read clearly as hidden. If you want the area to be truly unreadable, no block size is a substitute for a solid fill.

## Your picture stays in this browser tab

Opening a picture here does not send it anywhere. The page reads the file with your browser's file API, draws it on a canvas inside the tab, and builds the exported copy in memory. The download link points at that in-memory copy (an address that starts with `blob:`), not at a server. Close the tab and the copy is gone. This website consists only of static files; there is no upload address for a picture to go to.

## Under the hood

The editor works on a scaled preview for speed and records each box in the original picture's coordinates. On export it paints the full-size image, applies the chosen fill to each box, and encodes the result as a new file, in a background worker where the browser supports it. To confirm that no picture leaves your device, keep the developer tools' Network panel open during an export; [how to check this yourself](/methodology/) walks through it step by step in Chrome, Firefox and Safari.

## FAQ


### Can someone un-pixelate my image?

Sometimes. A 2016 study trained a neural network that picked the right person out of 530 in 57% of faces pixelated with 16 × 16 blocks, where blind guessing scores 0.19%. For text, the Depix and Unredacter projects rebuild pixelated words when the font and block grid can be guessed. For anything sensitive, use Solid block.

### Is a bigger mosaic block safer?

Bigger blocks remove more detail, so they help against casual viewing. They do not change the basic weakness: each block still holds an average of the real pixels. A solid fill holds none, which is why it is the default.

### Does the exported picture still have my location in it?

No. Export builds a new image and does not copy EXIF, GPS, XMP, IPTC or Content Credentials into it. If you only want the location removed and no pixels changed, use the location remover instead.

### Why does the editor show a smaller version of my photo?

Huge photos make a browser slow, especially on phones. The preview is scaled down to roughly 2,048 pixels at its widest, but your boxes are scaled back up and applied to the full-size picture when you export.

### Can I pixelate an image on my iPhone?

Yes, in Safari or any other browser on the phone. Safari can also open HEIC photos directly. After export, the share button lets you send the new copy to Photos or another app.

### What happens if the export fails on a very large picture?

The tool makes a second attempt at 4,096 pixels across the larger dimension and shows a note saying the copy was made smaller. The covering is applied exactly the same way.

## Sources

Checked 2026-10-11.

- [McPherson, Shokri and Shmatikov, 2016 (arXiv 1609.00408)](https://arxiv.org/abs/1609.00408): face recognition results against mosaics, including 57% top-1 at 16 × 16 windows.
- [Depix project README](https://codeberg.org/spipm/Depixelization_poc/): recovering text from screenshots pixelated with a linear box filter.
- [Bishop Fox, "Never, Ever, Ever Use Pixelation for Redacting Text" (2022)](https://bishopfox.com/blog/unredacter-tool-never-pixelation): the Unredacter tool and the advice to use solid bars.

Last updated: 2026-10-17