Background removal without a subscription

You need a product shot on white, or your face without the kitchen behind it. Every result is a website that wants an upload and a card. Background removal is separating the subject of a photo from everything behind it and making that part transparent, and your phone and your operating system already do it for free.

The paid background removal tools are not a scam. They are better at hair, glass and motion blur. For the other eighty per cent of pictures, the free version on the device you are holding is the same picture.

The one already on your device

Try this before opening a browser, because it takes four seconds.

On iPhone and iPad, open a photo, press and hold on the subject. It lifts away from the background with an outline. Drag it into a message, or tap Copy and paste it into anything that takes an image. Nothing is uploaded, and it works in Photos, Safari and Files.

On a Mac, right-click an image in Finder or Preview and look for Quick Actions, then Remove Background. It writes a new PNG next to the original.

On Android, Google Photos has the equivalent through its editing tools, and Samsung devices have the same press-and-hold gesture as iOS.

Three things to know about built-in background removal:

• It outputs PNG. That preserves transparency; a JPEG cannot, which is why saving as JPEG puts a white box back.
• It runs locally, so a photo of a person, a document or a product you have not launched is not leaving your device.
• It has no controls. You get its answer, and if the edge is wrong there is nothing to adjust.

That last point is the whole reason paid tools exist.

Judge the result on the edges, not the middle. Every tool gets the middle right.

Why does the edge look wrong?

Background removal fails at the edge for four reasons, and three of them are the photograph rather than the software.

The subject and the background are the same brightness. A grey jacket against a grey wall gives the algorithm nothing to separate. This is the single most common failure and it is fixed when the photo is taken, not afterwards.

The edge is genuinely ambiguous. Hair, fur, feathers, tulle. There is no correct answer at the pixel level, only a plausible one, and different tools guess differently.

The subject is moving. Motion blur means the boundary spans several pixels of mixed colour, and any cut through it looks wrong.

There is coloured light spilling. Standing next to a red wall puts a red rim on the subject that follows them into the cutout, where it looks like an error.

The last one has a name in production work — it is why green screens are lit carefully — and it is the one people mistake for a software fault. Nothing removes it automatically; it is repainted by hand or avoided when shooting.

Shooting so it works the first time

If you control the photograph, five seconds of setup beats any amount of correction afterwards. Background removal is only as good as the edge it was given.

1. Put contrast behind the subject. Not necessarily white — just clearly different in brightness from the subject's edges.
2. Separate them. A metre between subject and wall removes shadows that read as part of the outline.
3. Use even, diffuse light. Hard light creates hard shadows, which get cut out with the subject.
4. Avoid coloured surfaces nearby, for the spill reason above.
5. Shoot slightly wider than you need, so the edge is not against the frame border.

Point 2 is worth more than the other four together. A subject pressed against a wall has a shadow line that no tool can distinguish from the subject, because visually it is not distinguishable.

What to do with the cutout afterwards

A transparent PNG is a starting point, and two decisions follow immediately.

Which format to save. Keep PNG if the transparency matters — a logo, an overlay, anything going on top of something else. Convert to JPEG only when you have flattened it onto a solid background, because JPEG has no alpha channel and will fill the transparency with white or black.

What size. A cutout from a phone photo is large, and a 4,000-pixel-wide PNG for a 400-pixel avatar is a slow page and a big file. Resize after cutting, not before — the algorithm has more to work with at full resolution.

If the cutout is going into a video or a slideshow rather than a document, the alpha channel usually does not survive the round trip; most video formats do not carry transparency, so the image needs flattening onto its final background first. Building that kind of composition is covered in video tools, and the compression decisions in image tools.

Checking the result before you use it

A cutout that looks right in the editor can look wrong in place, and there are three quick checks that catch almost everything.

View it against the opposite tone. A cutout made on a white canvas hides white fringing completely. Drop it onto a dark background for two seconds and any leftover halo becomes obvious immediately. Then check it against white again for the same reason in reverse.

Zoom to one hundred per cent at the edges, not the middle. Every tool gets the middle right, so the middle tells you nothing. Follow the outline where the subject meets the old background, and look specifically at concave corners — under an arm, between fingers, inside a handle — because that is where background removal leaves fragments of the original scene behind.

Look for what got cut that should have stayed. A thin strap, a phone cable, an antenna, the stem of a glass. Thin features are the ones models drop, and you will not notice from a thumbnail.

If a cutout fails these checks and the photo cannot be retaken, the repair is manual: an eraser at low hardness along the offending edge, or a slight contraction of the selection by one or two pixels to pull the outline inside the fringe. Both are faster than trying a fourth tool on a photograph that was always going to be difficult.

When is a paid tool actually worth it?

Paid background removal earns its price in three cases, and none of them is "I need one photo done well".

Volume. Two hundred product images is a batch job, and doing it one long-press at a time is an afternoon. This is the strongest reason by a wide margin.

Hair against complexity. Portrait work against real backgrounds is where the specialist models are genuinely ahead, and the difference is visible at normal viewing size rather than only when zoomed.

A repeatable pipeline. If the same operation happens every week to the same kind of image, a tool with presets and an API saves more than it costs.

For a single image, the honest answer is that the free tool on your phone produces the same result, and the ten minutes spent finding a website was the expensive part. Remove.bg publishes its API documentation if you reach the volume case, which is the point at which paying starts to make sense.

One caution that applies to every uploaded alternative: a photo of a person, an unreleased product or a document is data. A free web tool has to be paid for somehow, and the terms usually permit retention. For anything you would not post publicly, use the version that runs on your device — which is the same advice as online tools generally.

The short version

Press and hold the subject. Save as PNG. Resize afterwards. That is background removal done, and it covers most of what people open a website for.

Reach for a paid tool when you have volume, hair against a busy background, or a weekly pipeline — and shoot with contrast behind the subject, because the best background removal is the one that had an easy edge to find.