About Background Remover
Most free background removers upload your photo to a server running a segmentation model. That works well on people, costs the operator money on every image, and means your photo — which might be a passport picture or a product you have not launched — sits on someone else’s machine. This one runs entirely in your browser and never sends the file anywhere.
The trade is honest: without a model, this cannot guess where a person ends and a cluttered living room begins. What it does very well is any subject on a reasonably plain background — product shots, logos, scanned documents, headshots on a backdrop, and anything filmed against a green or blue screen. For those, a well-tuned colour cut-out is often sharper than a model’s output, because it follows the actual pixel boundary rather than a predicted one.
The important detail is how background is identified. The naive approach removes every pixel matching the colour, which punches holes through the whites inside your subject: eyes, teeth, a shirt, the highlight on a bottle. The default here instead spreads inward from the border and stops when it meets the subject, so an enclosed white region is never reached — it cannot get there without crossing something that is not background.
Edges are graded rather than switched. A hard threshold produces the jagged outline that instantly reads as a bad cut-out, because real photographic edges are antialiased and a boundary pixel is a genuine blend of subject and background. Alpha fades across a band you control, which is what makes the result sit convincingly on a new background.
How to remove a background
Add your image
The border colour is sampled automatically, so the first result is usually close already.
Click the actual background
If the guess was wrong, click the background in the preview to sample that colour instead.
Tune tolerance and softness
Raise tolerance until the background clears; add softness until the edge stops looking cut out.
Download a PNG
PNG preserves transparency. JPG cannot — saving as JPG flattens the result onto a solid colour.
Which method to use
- Connected edges — the default, and right for nearly every photograph. Removes only background reachable from the border, so light areas inside the subject survive.
- Every matching pixel — right for a green or blue screen, where the key colour genuinely should vanish everywhere it appears.
- Turn on spill removal for green and blue screens. It neutralises the coloured rim the screen reflects onto your subject, which survives keying because it is not quite green enough to be cut.
When this will not work well
A busy or gradient background defeats any colour-based method, because there is no single colour to key on. Hair, fur and motion blur are the other hard cases: they are genuinely semi-transparent across many pixels, and no threshold reproduces that convincingly.
For those, a model-based remover really is better, and it is worth saying so rather than pretending otherwise. For a product on a white sweep or anything shot against a screen, this will match or beat one — and it does it without your image leaving the device.