BrowserTools

Background Remover

AI background removal, fully on-device

By Guga — BrowserTools maintainer, Tbilisi, Georgia Updated 2026-09-04
Free Private No uploads

The AI tab downloads a segmentation model — tens of megabytes — from imgly's CDN the first time you use it; your browser caches it afterwards, and no image is ever uploaded. The Color key tab is pure canvas and loads nothing at all.

What is Background Remover?

Two methods, and they suit different problems. The AI tab runs a segmentation model on your own machine, which is what you want for a person, a product with soft edges, hair, or anything with a genuinely complicated background. The Color key tab reads the pixel in the corner of the image and makes every pixel within a tolerance of that colour transparent, which is faster, exact, and much better for screenshots, logos, and studio shots against a flat sweep. Both produce a PNG with a real alpha channel, previewed against a checkerboard so you can see the cut. Be aware of the first run: the model is tens of megabytes and is fetched from a CDN on initial use, then cached by the browser. Your images never leave the device either way.

Remove image backgrounds using an AI model that runs entirely inside your browser — no uploads, no API keys, no watermarks. Also includes an instant color-key mode for solid backgrounds.

Private Background Remover In Your Browser

Background Remover is processed locally with client-side JavaScript. Your files never touch a server, and there is no queue to wait in.

Step-by-Step

How to use Background Remover

1

Pick a tab: AI removal (best quality) for subjects, or Color key (instant) for flat backgrounds.

2

Drop the image into the panel — on the Color key tab, the corner pixel is read automatically as the key colour.

3

On the Color key tab, set Tolerance to widen or narrow the match and adjust the background colour if the auto-detect is wrong.

4

Press Remove background or Remove background color, check the checkerboard preview, then download the PNG.

Key Capabilities
  • Two independent methods: an on-device segmentation model and an instant colour-key pass
  • The colour key is auto-detected from the image's top-left pixel and can be overridden with a colour picker
  • Tolerance runs from 2% to 90% in 2% steps, controlling how far a pixel may stray from the key colour
  • Both modes output a PNG with a genuine alpha channel, previewed on a transparency checkerboard
Common Use Cases
  • A marketplace seller cutting a cluttered product photo down to a clean cut-out for a listing.
  • A designer preparing a headshot with a transparent background for a team page layout.
  • A developer removing a flat white background from a UI screenshot so it can sit on a coloured page.
Clear Answers

Frequently Asked Questions

Is the first run slow because of a download?
Yes. The AI tab fetches a segmentation model of tens of megabytes from imgly's CDN on first use, and the progress bar tracks it. Your browser caches the model, so subsequent removals start immediately. The Color key tab downloads nothing and is ready straight away.
Which mode should I use?
Flat, uniform background — a screenshot, a logo, a product on a white sweep — use Color key; it is exact and instant. Hair, fur, translucent edges, or a busy scene need the AI tab, because those edges cannot be resolved by a colour threshold.
Why did the colour key eat part of my subject?
Tolerance is measured as distance in RGB space, so a white background and a white shirt are the same colour to the algorithm. Lower the tolerance to catch only near-exact matches, or switch to the AI tab. Raise it only when the background is noisy or slightly off-white.
Are my images uploaded to be processed?
No. Inference runs through WebAssembly in the tab, and colour keying is plain canvas pixel manipulation. Only the model file itself is downloaded on the first AI run; the image is never transmitted.