Background Remover

Background Remover

Cut out people, products, pets and objects from JPG, PNG or WebP photos. A small AI segmentation model runs on your device, so you get a transparent PNG — or a white or custom-colour background — without uploading anything.

Back

1. Image

Local · not uploaded

Drop a photo here

or click to browse

JPG · PNG · WEBP · BMP · AVIF

The AI model (about 4.4 MB) is downloaded from this site the first time you use the tool and runs on your device. Works best with one clear subject — a person, product, pet or object.

2. Background & edges

Updates the result
New background
Add an image to remove its background.

3. Before & after

Drag to compare

The original and the result appear here.

OriginalResult

Results can be imperfect around hair, fur, transparent or reflective objects, fine edges, shadows and busy backgrounds. Always check the cut-out before using it.

About this tool

About Background Remover

Background Remover separates the main subject of a photo from its background using U²-Net (u2netp), an open-source salient-object segmentation model. The model and its ONNX Runtime WebAssembly engine are bundled with this site and run in your browser. The prediction becomes a real alpha channel, so you get a transparent PNG — or a white or custom-colour background as PNG or JPG — with a draggable before/after comparison.

Product photos, profile pictures, stickers, presentations and marketplace listings often need the subject without its background. Doing it by hand takes time, and many online removers upload your photo to a server. Here the image stays on your device. The model is small, so results are good for clear subjects but not perfect: hair, fur, glass, shadows and busy or similar-coloured backgrounds often need touching up.

How to use

Step by step

  1. Drop a JPG, PNG or WebP photo onto the upload area.
  2. Wait while the model downloads from this site (first use only) and runs on your device.
  3. Drag the comparison slider to check the cut-out against the original.
  4. Choose a transparent, white or custom-colour background.
  5. Adjust Edge style and Edge smoothing if the outline looks too hard or too soft.
  6. Download the PNG, or the JPG when a solid background is selected.
Key features

What it offers

  • Bundled AI segmentation model (U²-Net, u2netp) — no server, no API key
  • Real transparent PNG output with an alpha channel
  • White or custom-colour background, with PNG or JPG download
  • Edge style (soft, balanced, crisp) and edge smoothing in pixels
  • Draggable before/after comparison slider, also keyboard-accessible
  • Original size, output size and dimensions shown
  • EXIF rotation applied so phone photos stay upright
  • Clear messages for empty, damaged, unsupported, transparent-only and oversized images
Best practices

Tips for best results

  • Use photos with one clear subject that stands out from the background.
  • Good lighting and contrast between subject and background give cleaner edges.
  • Zoom in on hair, fur and fine details before using the result.
  • Use slight edge smoothing for soft subjects and Crisp for products with hard edges.
  • Keep the original photo — the model can miss parts of complex subjects.
Benefits

Why people use it

Private

Photos are processed on your device and never uploaded.

Fast

Results in seconds once the model is loaded.

Flexible

Transparent, white or coloured backgrounds.

Honest

Edges are shown side by side so you can judge the result.

FAQ

Common questions

No. The image is decoded and processed in your browser. The AI model and its WebAssembly engine are downloaded from this website once and run on your device.
It uses U²-Net (the small “u2netp” version), an open-source salient-object segmentation model. It predicts which pixels belong to the main subject; that prediction becomes the transparency of the output PNG.
Not always. Hair, fur, glass, transparent or reflective objects, fine details, soft shadows and backgrounds similar in colour to the subject are often imperfect. The model works on a 320 × 320 version of the image, so very fine detail can be lost; the edge style and smoothing controls help, but check every result.
Edge style changes how sharply the model’s mask switches between subject and background: Soft keeps its gradual edge, Crisp makes it harder. Edge smoothing feathers the mask edge by the chosen number of pixels.
Yes. Choose White or Colour to fill the background. You can then download a PNG or a smaller JPG. With a transparent background only PNG is offered, because JPG cannot store transparency.
Files up to 30 MB and images up to 50 megapixels can be opened. Images above 16 megapixels are scaled down to about 16 megapixels to stay within browser memory, and you are told when this happens.
Current versions of Chrome, Edge, Firefox and Safari that support WebAssembly and module workers. The first run takes longer while the model downloads; later runs reuse the browser cache.

Related tools

Image Cropper

Crop freely or to fixed ratios (1:1, 16:9, 4:3, 3:2) with a live drag overlay.

Launch tool

Image Format Converter

Convert between PNG, JPG, WebP, BMP, and more, with quality control where supported.

Launch tool

Image Compressor

Shrink PNG, JPG, and WebP with a live quality slider and before/after size readout.

Launch tool
Private by design

This runs entirely in your browser

No file or text you enter here is uploaded. Encoding and decoding happen on your device using native browser APIs — close the tab and nothing remains on a server.

Install Toolisco Add it to your device for offline, one-tap access.