Background Remover
Cut the subject out of a photo in one click. The AI model runs entirely in your browser, so the picture is never uploaded, and you get a transparent PNG ready to use.
- Files stay on your device
- No sign-up
- Free to use
How to use Background Remover
- Drop a photo onto the upload area. Pictures with a clear main subject work best.
- Wait a few seconds while the model loads and finds the subject. The first run takes longest.
- Compare the result with the original using the slider, and choose a transparent, white or custom-colour background.
- Download the cut-out as a PNG.
Background Remover features
Automatic subject detection
A neural network identifies the main subject without any manual tracing.
On-device AI
The model is downloaded once and runs in your browser, so your photos stay private.
Transparent or solid background
Export with transparency, or drop the subject onto white or any colour for product shots and ID-style photos.
Edge softening control
Adjust how hard or soft the cut-out edge is.
Before/after comparison
Drag the divider to check the cut-out against the original.
Original resolution
The output has the same pixel dimensions as your photo.
When to use Background Remover
- Creating product images on a clean white background for an online shop.
- Making a profile photo or headshot with a plain backdrop.
- Cutting out a subject to place into a presentation, collage or thumbnail.
- Preparing stickers, logos or design assets with transparency.
Background Remover FAQ
What kinds of images work best?
Photos with one clear subject that contrasts with the background: a person, a product, an animal, a vehicle. Busy scenes, several overlapping subjects, or a subject the same colour as the background give less accurate results.
How well does it handle hair and fine details?
The model produces a soft mask, so hair outlines are approximated rather than traced strand by strand. For studio-grade cut-outs of fine hair or transparent objects, a manual editor will still do better.
Is my photo uploaded?
No. The model file is downloaded to your browser and all processing happens on your device.
Why is the first run slow?
The model and its runtime, around 20 MB in total, are downloaded and initialised the first time. They are cached afterwards, and each image then takes roughly one to five seconds depending on your device.
What format is the result?
PNG, which supports transparency. If you choose a solid background colour you can convert the result to JPG with the PNG to JPG tool to make it smaller.
How automatic background removal works
Background removal is a segmentation problem: for every pixel, decide whether it belongs to the subject. This tool uses U²-Net, a neural network designed for salient object detection, in its compact form. It was trained on thousands of images with hand-drawn masks and learned to recognise what the eye would consider the main object in a picture.
The image is scaled to the size the network expects and passed through it using ONNX Runtime compiled to WebAssembly. The output is a greyscale map in which bright areas are subject and dark areas are background. That map is scaled back up to your image's resolution and used as the alpha channel, so the subject stays opaque and the background becomes transparent.
Because the mask is predicted at a modest resolution, edges are smooth rather than pixel-exact. The softening control lets you tighten the edge for hard-edged products or relax it for hair and fur.