Remove a solid background
Make a flat background transparent. Built for green screens, product shots and logos, not for photographs of people.
What this does: clears a background that is one flat colour, by spreading inward from the edges of the picture. Green screens, products on white, logos, scanned signatures. What it does not: separate a person or an object from an ordinary photograph. That needs a trained model and this deliberately has none, because a model would be tens of megabytes and most tools that offer it are uploading your picture instead.
No image yet.
What it actually does
It starts at every pixel around the border of the picture and spreads inward, clearing anything close enough in colour to where it started. When it reaches something different enough, it stops. What survives is whatever is not connected to the edge by a run of similar colour.
That description is also the list of its limits, which is why it is worth reading. Connectivity from the edge is the whole mechanism, so a background that varies across the frame will not clear evenly, and a gap fully enclosed by the subject, such as the space inside a handle, is not connected to the edge and will not clear at all.
What it is genuinely good at
Product shots on white. The single most useful case: a photograph taken on a white sheet or in a lightbox comes out cleanly, ready to drop onto a page or a listing.
Green screens. What chroma key was invented for, and it works exactly as you would expect.
Logos and graphics. A logo supplied as a JPG on white, which happens constantly, converts back to a transparent PNG in one pass.
Scanned signatures. Sign a white sheet, photograph it, clear the paper, and you have a signature you can place on a document.
Setting the tolerance
Low values clear only the exact shade, which leaves a halo on a backdrop that is slightly uneven, and real backdrops always are. High values clear more, and eventually start eating into the subject wherever its colours approach the background.
Start at 30. If a fringe remains, raise it. If the subject starts dissolving, lower it. The figure on the right of the slider tells you how much of the picture was cleared, which is a quicker guide than squinting: a product on white usually lands somewhere between 40 and 70%, and a sudden jump to 95% means it has broken through into the subject.
Why there is no model here
Removing the background from an arbitrary photograph is a genuinely hard problem and the tools that do it well run a neural network trained for the job. That is a real answer, and it is not this one.
Shipping one would mean tens of megabytes downloaded before anyone could press anything, on pages built to be small. The commoner alternative is worse: most free background removers upload your photograph to a server and run the model there, which is exactly the arrangement these pages exist to avoid. Doing the narrow job honestly beats doing the broad job by quietly sending your pictures away.
Save it as PNG
JPG has no transparency, so saving a cut-out as JPG fills everything you just removed with white. Only PNG and WebP are offered here for that reason. WebP is meaningfully smaller and every current browser displays it; PNG is the safer choice if the file is going into older software.
Common questions
- Will this work on a photo of a person?
- No, and it is better to say so than to waste your time. This works by spreading inward from the edges of the picture and clearing anything close to the colour it starts on, so it needs a background that is one flat colour: a green screen, a product on white, a logo. A person photographed in a room or outdoors has a background that varies from pixel to pixel, and no amount of tolerance separates it cleanly. That job genuinely needs a trained model, which this deliberately does not have.
- Why not use AI like the other tools do?
- Because a segmentation model is tens of megabytes that would have to download before you could do anything, and these pages are built to be small and to run without a server. A tool that removes the background from an arbitrary photograph is either running that model on your machine or, far more often, uploading your picture to somebody else's. This one does neither, and is honest about the narrower job it does instead.
- What does the tolerance do?
- It sets how different from the starting colour a pixel can be and still count as background. Low values clear only the exact shade, which leaves a fringe on a slightly uneven backdrop. High values clear more, and eventually start eating into the subject where its colours approach the background. Start around 30 and move it until the edge looks right.
- Why must I save as PNG?
- Because JPG has no transparency at all. Saving a knocked-out image as JPG would fill everything you just removed with white, which defeats the exercise. PNG and WebP both keep the transparency, and WebP files are smaller.
- Is my image uploaded?
- No. It is read from your own disk, processed in your browser, and handed straight back. Nothing is sent to us or to anyone else, and nothing is stored.