How it works : Mask an area
This tool does not blur, and that is deliberate: blurring and pixelation both come undone. It replaces the pixels with an opaque fill, the only method of which it can be said that nothing is left to recover.
This workshop's plan announced a blur tool. Checking the state of the art changed the promise, and saying so is more honest than shipping something reassuring. A blur COMES UNDONE: published work recognises faces and reads blurred text at high rates. Pixelation is weaker still as soon as the space of possibilities is small: for a number, an identifier, a plate or an amount in a known font, the original string is recovered by comparing mosaics one by one.
That is why data protection authorities recommend neither, for removing information from an image, but an OPAQUE COLOUR FILL. So this tool applies a black fill by default, directly onto the pixels. Pixelation is still offered, because it is sometimes wanted for the look, and it is labelled as less safe BEFORE you click.
One last trap, and it has cost real publications dearly: a JPEG often carries a preview thumbnail in its metadata, and masking the pixels without touching the container leaves that ORIGINAL thumbnail in the file. Here the image is re-encoded from a canvas: the thumbnail no longer exists, nor does any other metadata.
How it works
- Drop your image in, then draw an area straight onto it, with a finger, a mouse or the keyboard.
- Choose the opaque fill (recommended) and click “Apply this area”. The area is masked straight away, on the displayed image: what you see is what you will get. Repeat as many times as needed.
- Click “Save the masked image”. The tool tells you whether ALL areas are irreversible, or whether one stays weak.
Formats accepted
Any format your browser can display can be masked. The output is PNG or JPEG: those are the two formats every browser can write.
PNG
The right choice for a screenshot: nothing is lost and text stays crisp.
JPEG
Lighter on a photograph. The original file's preview thumbnail disappears in the re-encode, which is essential here.
WebP
Read everywhere, accepted as input. The output does not offer WebP: no version of Safari can write it.
AVIF
Accepted as input if your browser displays it. No browser writes it provably.
Its limits
There is no blur in this tool, and that is a decision
A blur is a largely invertible mathematical operation: it can be undone by deconvolution, and trained models read blurred text. Offering a blur while promising anonymity would be a comfortable lie. We prefer a tool that really masks and says why.
Pixelation is offered but flagged as less safe
It keeps the average colour of each block, so an exploitable trace. For text in a known font, the string is recovered by comparing mosaics against those of every candidate string: very effective on a number or an identifier. It stays useful for the look, and the tool draws the distinction clearly.
The image is re-encoded, so the original preview thumbnail disappears
This is a limit that is in fact a protection. A camera JPEG often contains a small copy of the photo in its metadata; masking that only rewrote the pixels would leave that UNMASKED copy in the file. Re-encoding from a canvas removes it, along with the date, the camera and the GPS position.
16.7 million pixels maximum
A browser canvas has a ceiling, and the compatibility floor is 16.7 million pixels (iPhones before iOS 18). Beyond that the operation fails silently and returns a blank image: the tool refuses cleanly, showing the dimensions it read.
The masking is permanent in the produced file
That is exactly the point: if it were reversible it would mask nothing. Your original file is untouched, it stays on your device. Keep it if you still need it.
Questions about this tool
Is my image sent anywhere?
No, and it is checkable in thirty seconds. Open the inspector with the F12 key, “Network” tab, clear the list, then drop the image in and mask. Not a single line appears. That matters especially here: what people mask is precisely what they want nobody to see.
Why not offer a blur? Everyone else does.
Because a blur does not mask. It is a largely invertible operation: it is undone by deconvolution, and trained models recognise faces and read blurred text at high rates. Data protection authorities recommend an opaque colour fill for removing information from an image, and that is what this tool does. A blur that promises anonymity is false comfort, and this site exists not to sell any.
What about pixelation, then?
It is here, but labelled as less safe, and the tool says so again after the act if you used it. It keeps the average colour of each block: on text in a known font, one can build the mosaic of every possible string and find the right one. On a number, an identifier, a plate or an amount, that is very effective. Use it for the look, not to protect somebody.
Is it really irreversible?
With the opaque fill, yes, in the produced file: the original pixels are no longer in it, there is no layer to peel off and no average to exploit. That differs from a black rectangle placed on a PDF, which remains an object drawn ON TOP of text that is still selectable. Your original file is untouched: keep it elsewhere if you need it.
Why do you talk about a preview thumbnail?
Because it is a real trap that has already leaked images. A camera often slips a small copy of the photo into the file's metadata. A tool that masked the pixels without rewriting the container would leave that unmasked copy inside, and anyone could extract it. Here the image is re-encoded from a canvas: the thumbnail no longer exists, and the site's automatic check verifies that on the produced file.
Can I mask several areas?
As many as you like. Each area is applied immediately to the displayed image, and the table keeps the list with the method used for each. What you see on screen is exactly what the file will contain.
This tool does not blur, and that is deliberate: blurring and pixelation both come undone. It replaces the pixels with an opaque fill, the only method of which it can be said that nothing is left to recover.
VerifyThe proof, step by stepThe network inspector check, explained one gesture at a time, with screenshots.
The cornerThe image cornerA photo says more than it shows. Here you see it, clean it, slim it down, crop it, and blur what is nobody's business.