Removing image backgrounds — in your browser, no upload, no watermark

Product photos for a listing, a profile picture without your messy room, a logo that needs to sit on any colour — background removal is one of the most-wanted image edits, and the popular web services meter it, watermark it, or make you sign up.

How it works here (and why it's private)

Remove Background runs a machine-learning segmentation model inside your browser. The first run downloads the model (a few MB), then everything happens on your machine: no upload, no queue, no per-image credits, no watermark. The output is a PNG with true transparency.

What it's good and bad at

ML segmentation shines on clear subjects: people, products on contrasting backgrounds, animals, cars. It struggles where humans also squint:

• Hair and fur against busy backgrounds — expect slightly soft edges.
• Glass and transparency — a bottle's edges confuse any segmenter.
• Subject-coloured backgrounds — a white shirt on a white wall gives the model little to work with.

The best fix is at capture time: photograph the product against any plain, differently coloured surface, in even light. Even a bedsheet works.

After the cut-out

• Marketplace listings (Amazon-style) usually demand a pure white background: put the transparent PNG on white and export — or keep transparency for design work.
• Profile pictures: crop the cut-out into a circle with Circle Crop.
• Cards and banners: Add Border or drop the PNG straight into the Design Studio on any canvas.

PNG vs JPG warning

Save cut-outs as PNG. The moment you convert to JPG, transparency is replaced with a solid background (usually white) — JPG simply has no alpha channel. That's not a bug in any tool; it's the format.