Make an image background transparent
Drop in an image and the tool clears the colour behind your subject. It removes a colour rather than cutting out a person, so it works best on a plain background.
The pixels are read and rewritten inside this page, so the image stays on your device.
How to use it#
- Drop an image onto the box above, or click it to choose one.
- Check the colour it picked, or click anywhere in the preview to choose a different one.
- Raise the tolerance until the background is gone, and the edge softness until the outline stops looking cut out.
- Download the result as a PNG.
What this removes, and what it does not#
This clears pixels by colour. You tell it which colour the background is, it measures how close every pixel is to that colour, and anything close enough becomes transparent. On a product shot against a white sweep, a logo on a solid field, or a scan of a signature, that is exactly the right tool and it finishes instantly.
It cannot tell a person from a wall. If your subject sits in front of a room, a street or a sky with clouds in it, no colour threshold will separate them, and pushing the tolerance up just eats holes in the subject. That job needs a trained model reading the image, which is a different kind of software and not something that can run privately in a browser tab.
One more limit worth knowing: any colour in your subject that matches the background gets removed too. White text on a white background disappears along with it, which is usually visible straight away in the preview.
Tolerance and edge softness#
Tolerance is how far a pixel can be from your chosen colour and still count as background. Too low and you keep a halo of nearly-background pixels. Too high and the subject starts dissolving, usually beginning with its lightest or darkest parts.
Edge softness handles the pixels in between. A photograph has no hard boundary: the edge of an object is a band of pixels that are part subject and part background, which is what makes it look smooth. Clearing those outright leaves a jagged outline, and keeping them leaves a coloured fringe. Instead, this fades their transparency across a band past the tolerance, so an edge pixel that is three quarters background ends up three quarters transparent.
A useful order is to raise the tolerance until the background is gone, then raise the softness until the outline stops looking cut out. Both re-apply to the preview as you move them.
The image that only looks transparent#
A common frustration: you download something that showed a grey and white checkerboard behind it, and the checkerboard is still there in the file. Image editors draw that pattern to mean "nothing here", and a screenshot or a flattened export turns it into real pixels.
The tool spots that case when you drop such a file in and switches to the checkerboard setting, which removes both tones of the pattern at once instead of making you clear them one at a time. The result is the transparency the image was pretending to have.
Why the download is always a PNG#
Transparency needs a format with an alpha channel. JPG has none, so saving a cut-out as a JPG puts the background straight back, usually as a black or white box. That single fact causes more confusion than anything else on this page.
PNG is the safe answer and is what you get here whatever you dropped in. WebP also stores transparency and makes a smaller file, so it is worth converting the PNG afterwards if the image is going onto a website.
Frequently asked questions#
Is my image uploaded anywhere?
#
No. Every pixel is read and rewritten in this tab by your own browser. The image never goes to a server, and the tool carries on working with the network off.
Can it cut a person out of a busy photo?
#
No. It removes a colour rather than recognising a subject, so it needs a plain background to work against. Separating a person from a real scene takes a trained model, which cannot run in a browser tab without a very large download.
What does tolerance change?
#
How different from your chosen colour a pixel can be and still be cleared. Raise it when a rim of background survives around the subject, lower it when parts of the subject start disappearing.
Why does the cut-out have a fringe?
#
The pixels along an edge are a blend of the subject and the background, so clearing only exact matches leaves the blended ones behind as a coloured rim. Raising the edge softness fades those pixels out by how close to the background they are, which removes the rim without leaving a jagged outline.
What is the checkerboard setting for?
#
Images saved from an editor sometimes have its grey and white checkerboard baked in, so they look transparent and are not. That setting clears both tones of the pattern at once. The tool switches to it by itself when it finds a checkerboard around the edges of your image.
Why is the download a PNG when I gave it a JPG?
#
JPG cannot store transparency at all, so saving the result as one would fill the background back in. PNG has an alpha channel and is read by everything, so it is what the tool writes.