Background Removal
The Background tab cuts a subject out of its background using a real image segmentation neural network, run entirely in your browser.
How it works
The tool uses U2Net, an open-source segmentation model, running locally
via WebAssembly (through onnxruntime-web). Nothing is uploaded — the model
downloads once to your browser and runs against your image on your device.
First use
The first time you use Background removal, your browser downloads the U2Net model (roughly 200MB) and caches it for future use. This means:
- The very first run takes longer, since it includes the download
- Every run after that is fast, since the model is already cached locally
Tip
If you're on a slow connection, start the first background-removal run ahead of time so the model has finished downloading before you actually need the result.
How to use it
- Open the Background tab
- Drop in a photo
- Wait for the model to load (first time only) and process the image
- Download the result with the background removed