Background Remover

Cut out people, products and pets from a photo and download a transparent PNG. The AI model runs on your own device — your photo is never uploaded.

Jump to result

1. Choose a photo

or drag & drop / paste a JPG, PNG or WebP (max 25 MB)

Works best for a clear subject — a person, product or animal — on a background that stands out from it. Fine hair, glass and busy scenes may need a touch-up with the refine brush.

2. Remove the background

The first run downloads the AI model (about 4.5 MB) and the AI engine once; after that it is much faster.

3. Refine edges

0 keeps the soft AI edge. Increase it for crisper, harder edges and to remove faint leftover background.

Refine brush

Open the Refine view of the result and paint with your mouse or finger. Erase removes leftover background; Restore brings back parts of the subject.

4. Background & download

Background for the downloaded image
Result

Your cut-out will appear here.

How it works

On-device AI

A compact image-segmentation model (U²-Net) runs inside your browser using WebAssembly. It looks at the whole photo and estimates, for every pixel, how likely it is to belong to the main subject.

Private by design

Your photo is processed in this tab only. Nothing is uploaded, stored or shared — you can even disconnect from the internet after the model has loaded.

Refine & export

Use edge hardness for crisper edges and the brush to erase or restore areas. Download a transparent PNG or place the subject on white or any colour.

Tips for the best cut-out

  1. Use a photo where the subject is in focus and clearly separated from the background.
  2. Plain or blurred backgrounds give the cleanest edges; similar colours (black hair on a dark wall) are harder.
  3. If a little background is left around the subject, raise edge hardness first, then erase the rest with the brush.
  4. For passport or ID photos choose the White background before downloading.

Frequently asked questions

Why is the first run slower?

The first time, your browser downloads the AI model (about 4.5 MB) and the WebAssembly engine, then prepares them. They are cached by your browser, so later images on this page — and usually later visits — are processed in a few seconds.

Is my photo uploaded to a server?

No. The model is downloaded to your browser and the whole cut-out is computed on your device. Your photo never leaves it.

Why are some edges or hair not perfect?

The model is small so it can run on phones. It is very good with people, products and animals, but fine hair strands, transparent objects and cluttered scenes can confuse it. Use edge hardness and the Erase / Restore brush to fix those areas.

What size is the downloaded image?

The PNG has the same size as your photo. Very large photos are scaled down so the longest side is at most 4096 pixels to keep your browser from running out of memory.

Which AI model is used?

U²-Net (the small “u2netp” version) by Xuebin Qin et al., released under the Apache-2.0 licence and packaged by the open-source rembg project. It runs with ONNX Runtime Web by Microsoft (MIT licence).

It says my browser is not supported — what can I do?

The tool needs WebAssembly, which every current version of Chrome, Edge, Firefox and Safari supports. Update your browser, or try a different one. Very old phones may also run out of memory with large photos.