MultiKit

Background Remover: Remove a Picture's Background in Your Browser

  • Free
  • No sign-up
  • Runs in your browser
  • Version 1.0.2
  • Last updated 2026-10-25
Report a problem

Drop a picture and the background is taken out, leaving the main subject on transparency. Change the background to white or a color, adjust the edge, and download a PNG. It runs in your browser, so the picture never leaves your device.

It works best on a picture with one clear subject, such as a person, a product or an animal, and a background that is different from it. Fine detail such as hair, fur and glass can come out rough, and a picture with several things or a busy background may need the controls below or may not work. Check the result before you use it.

Everything you enter is processed in your browser and is never sent to a server.

Browsers:Chrome91+Edge91+Firefox89+Safari16.4+Opera77+Samsung Internet16.0+Browsers

How to use

  1. Choose a picture, drop it on the box or paste one. The first time, the tool and its model load, about 19 MB, which your browser keeps for next time. Then the background is taken out in a few seconds.
  2. Compare the original with the result. The checkerboard behind the result shows what is transparent.
  3. Choose what goes behind the subject: transparent, white or a color. If edges look too tight or too wide, move How much to cut and Softness of the edge.
  4. Pick a format and download. PNG and WebP keep transparency. JPG puts the subject on the chosen color, or on white if the background is transparent.

Examples

A photo of a product on a table
→ The product on transparency, ready for a shop page
A clear subject against a different background is the easiest case.
A portrait with a plain wall behind
→ The person on white, for an ID or profile picture
Hair and fur are the hardest parts, so check the edge at full size.
A picture of several things, or with a busy background
→ It may take the wrong thing, or only part of it
The model looks for one main subject. Try the controls, or crop to the subject first with Crop Image and send it here.

How it works

The tool uses U²-Net, a neural network trained to find the most eye-catching object in a picture, in its small form, called u2netp, which is about 4.4 MB. The picture is squeezed into a 320 by 320 square, brought to the form the network was trained on, and given to the network, which answers with a 320 by 320 map of how much each point belongs to the main subject. The map is stretched so that its lowest value is 0 and its highest is 1, scaled up to the size of the picture, and turned into transparency. How much to cut sets the value at which a point is half kept, and Softness of the edge sets how gradual the change is.

The network runs through ONNX Runtime Web, a WebAssembly program, on one thread. The model and the program are served from this site, so nothing is fetched from another server and nothing of yours is sent.

This is a small model, and the result is a guess at what the main subject is. It was tested with pictures drawn for the test and with the controls, and it works best on a clear subject against a different background. Look at the result before you use it.

What pictures work best?

  • Good candidates: a person, product or pet with a clear outline and a background that differs in color or brightness.
  • Output: download a PNG with transparency; use the edge controls to trade a cleaner cut for retained hair and soft detail.
  • Privacy: the image and the U²-Net inference stay in this browser. The model files come from this site, not a third-party image service.
  • Limits: busy scenes, transparent objects, shadows, fine hair and a subject that blends into the background may need manual cleanup. This tool finds one visually dominant subject rather than letting you outline several objects.

Sources

Frequently asked questions

How do I remove the background from a picture?
Choose the picture and wait a few seconds. The tool finds the main subject and makes everything else transparent. Then pick transparent, white or a color for the background, adjust the edge if needed and download.
Is my picture uploaded?
No. The model and the program that runs it are loaded from this site and run in your browser, so your picture is processed on your device and never sent anywhere.
How good is the result?
It is good for pictures with one clear subject, such as a person, a product or an animal, against a background that differs from it. It is less good with hair, fur, glass and other fine or see-through parts, with several subjects, and with a busy background. The model is a small one, chosen so that it loads fast and runs in any browser, so it will not match the best paid services on hard pictures.
Why does the first picture take longer?
The first time, about 19 MB is loaded: the model and the program that runs it. The browser keeps them, so later pictures start at once. Taking out the background itself takes a few seconds on most devices.
How large a picture can I use?
Up to 12 million pixels, for example 4000 by 3000. The model looks at a 320 by 320 version of the picture and the result is scaled up to the full size, so edges of a very large picture are smoother than detailed.
How can I fix an edge that is too rough or too tight?
Use the two controls under Edge. How much to cut moves the edge in or out, and Softness of the edge blurs it. For a hard cut, set softness to 0. If the wrong thing was kept, try cropping to the subject first.
Can I use the result for commercial work?
The tool is free and the model, U²-Net, is released under the Apache License 2.0, which allows commercial use. You are still responsible for having the right to use the picture you put in.
Version history
  • 1.0.2 — Adds a practical decision guide covering best-fit uses, inputs and outputs, privacy, quality trade-offs and known limits.
  • 1.0.1 — Makes the search title and description state the supported task, key formats and privacy benefit more clearly.
  • 1.0.0 — First release.

Last updated: