How does this remove backgrounds without uploading my photo?
It runs a neural network called RMBG-1.4 directly in your browser using ONNX Runtime Web. The model weights download once from Hugging Face, then every cutout is computed locally on your device. Your photo is never sent to a server — you can confirm that in your browser's network tab.
Is it really free with no watermark?
Yes. There's no signup, no credit limit, and no watermark on the result. The output is a transparent PNG you can use anywhere.
What image formats can I use?
PNG, JPG, WebP, AVIF, GIF, BMP, and iPhone HEIC/HEIF are all accepted. The result is always a PNG so the transparency is preserved — JPG can't store an alpha channel.
How do I get the best cutout?
The model is trained on people, products, and animals, and does best when the subject is clearly separated from the background. Good lighting and focus help. Very fine detail like loose hair or thin wires is the hardest case for any background remover.
Why does the first run take a moment?
The first time you remove a background, the browser downloads the ~44 MB model. After that it's cached, so later cutouts start immediately. Processing itself takes a few seconds depending on your device.
Common use cases
- Product photos for shops and marketplaces.
- Profile pictures and avatars with a clean transparent edge.
- Cutouts for thumbnails, posters, and presentations.
- Prepping images before compositing onto a new background.