# Censor a photo

> To censor a photo safely, cover the area with an opaque block and save a new file, so nothing of the original is left underneath. A thin bar across the eyes is not enough to hide a face, and some phone editors have left cropped-out pixels inside saved files, which is why this tool always encodes a fresh image.

## What people usually need to censor

"Censor" covers a lot of small jobs. Before a photo goes on social media, into a group chat, onto a listing site or into a support ticket, people commonly hide:

- a house number, a street sign or the view from a window that pins down where they live;
- letters, parcels and delivery labels lying on a table;
- a laptop or TV screen showing email, messages or a work dashboard;
- a whiteboard, a calendar or a sticky note with names and plans;
- family photos on the fridge, kids' names on artwork, school letters;
- keys, a car key fob or an ID card left in view;
- a tattoo, a logo on clothing or a badge that identifies someone;
- an offensive word on a sign or a gesture you do not want to republish.

The tool does not need to know what the thing is. You draw a box, it gets covered.

## Scan the whole frame first

The things that leak information are rarely the subject of the photo. They sit at the edges, in reflections and in the background. Before you draw anything, zoom in and move across the picture in a fixed order (left to right, top to bottom) and ask of each area: does this show where I am, who someone is, or something I would not hand to a stranger?

Pay extra attention to glass and shiny surfaces. Windows, mirrors, sunglasses, oven doors and phone screens reflect whatever is opposite them, including faces and screens you thought were out of shot.

## How to censor a picture

1. Load the picture with the file button, or drag it onto the page.
2. Drag a box over the first thing to hide. Keep going until everything on your list is boxed; boxes can be moved, resized from the corner handle, or deleted.
3. If there are people in the shot, press **Find faces** to box them automatically, then fix anything the finder got wrong. For text such as email addresses or phone numbers, **Find text** can suggest boxes.
4. Keep **Cover with** set to **Solid block**. Pixelate and Blur are there for looks only and can be partly undone.
5. Press **Export**, check the new copy at full size, and save it.

## Why a bar over the eyes does not work

The classic black strip across the eyes hides less than it seems. Face shape, hairline, ears, mouth, skin marks and the rest of the photo stay visible, and both people and software use them. If someone must not be recognized, cover the whole head, or at least the whole face from hairline to chin.

The same goes for half measures with text: covering only the middle of a name or one line of an address usually lets a reader fill in the gaps. Cover the full item.

## Opaque covers versus effects

Effects like pixelation and blur keep a scrambled version of what was there. Security researchers at Bishop Fox released a tool in 2022 that rebuilds pixelated text, and their advice for redacting text was black bars over the whole of it, and nothing else. Faces can be similar: recognition software has matched blurred and pixelated faces far more often than chance.

A solid block has nothing underneath once the image is exported. That is why it is selected when the page loads, and why the two effect modes carry a "decorative only" label.

## Make sure the original is really gone

How you save matters as much as what you draw. In 2023, Google patched a flaw (CVE-2023-21036, nicknamed aCropalypse) in the Pixel phone's screenshot editor: when an edited image was saved over the original, the old file was not cut short, so the end of the original picture was still inside the file and could be partly recovered. Microsoft fixed the same kind of problem in the Windows Snipping Tool that year.

This tool avoids that class of problem by never editing your file in place. Export draws the picture and your covers onto a blank canvas and encodes a completely new image. Your original file on disk is not touched at all; keep it private or delete it as you see fit.

## What the new copy contains

- Your covers, burned into the pixels.
- None of the hidden photo data. Camera settings, capture dates, GPS coordinates, captions and credits, the manufacturer's private notes, preview thumbnails and Content Credentials all stay behind in the original.
- The resolution you started with. You edit on a lighter working copy (its longest side is near 2,048 px), and the covers are mapped onto the real image at export. When that is too much for the browser, the saved copy is shrunk so its longest side measures 4,096 px, and a notice appears.

If all you need is to strip the hidden data and leave the pixels alone, the [photo metadata remover](/remove-metadata-from-photo/) does that without re-encoding.

## Photos and devices this works with

- **Formats:** JPEG, PNG and WebP everywhere; HEIC only in browsers that display it natively (Safari), otherwise convert to JPG first.
- **Devices:** current Chrome, Edge, Firefox and Safari on computers, Android phones, iPhones and iPads. On a phone you draw with your finger and save with the share button.
- **Several photos:** choose more than one and handle them one after another.
- **Not supported:** video, animated images and PDF files.

## What happens to your photo

The page reads your photo into the browser tab, you edit it there, and the new copy is created in the tab's memory. No part of the picture is sent to this site or anyone else; the site has no server code that could receive it. Face and text finding use models that are downloaded to your browser when you first press those buttons, and they run on your device. When you close the tab, the page's copies are discarded.

## Technical note

Editing happens on a canvas. The finished image is produced by drawing the full-size photo, painting each box, and encoding the canvas into a new file, off the main thread when the browser allows it. The [methodology page](/methodology/) walks you through checking your browser's request log during the edit, so you can confirm for yourself that the photo stays put.

## FAQ


### What is the safest way to censor part of a photo?

Cover it with a fully opaque block and export a new file. That leaves no trace of the original pixels in the covered area, and the new file is made from the visible pixels alone, so the original's hidden photo data is not carried into it.

### Can censoring be undone?

A solid block exported as a new image cannot be peeled back, because the covered pixels are replaced. Pixelate and blur keep a scrambled version of the area, which can sometimes be reconstructed.

### Does this change my original photo?

No. The original file stays exactly as it was. You get a separate new copy with the covers applied.

### Is a black bar over the eyes enough?

Usually not. The rest of the face, the hair and the setting still identify people. Cover the whole face, or more, when someone must not be recognized.

### Can I censor words automatically?

Find text reads the image on your device and suggests boxes for card numbers, IBANs, phone numbers, email addresses, long runs of digits and QR-style codes. Ordinary words need a box drawn by hand.

### Will the censored photo still show where it was taken?

The exported copy is a new image made from what you see, so the GPS and other EXIF data of the original are not carried into it. Anything visible in the picture itself, like a street sign, still needs to be covered.

## Sources

Checked 2026-10-11.

- [NVD entry for CVE-2023-21036](https://nvd.nist.gov/vuln/detail/CVE-2023-21036): Pixel screenshot editor failed to truncate edited images.
- [NVD entry for CVE-2023-28303](https://nvd.nist.gov/vuln/detail/CVE-2023-28303): Windows Snipping Tool information disclosure.
- [Bishop Fox on Unredacter (February 2022)](https://bishopfox.com/blog/unredacter-tool-never-pixelation): pixelated text recovered; solid bars recommended.

Last updated: 2026-10-17