How to remove the background from a photo
- 1Drop one or more photos on the box, tap it to pick photos on your phone, or paste a copied image.
- 2Wait for the AI model to load the first time. Each photo is then cut out on your device, one after another.
- 3Drag across the preview to check the edges, choose a transparent, white or colored background and download the image, or all photos as a ZIP file.
How the AI finds the subject
The tool uses ISNet, a neural network for dichotomous image segmentation that was trained to separate the main object of a photo from everything behind it. It looks at a copy of your photo of up to 1024 pixels, gives every pixel a value between foreground and background, and that soft mask is scaled back to the full size of your photo. The values in between keep soft edges, so hair and fur fade out instead of ending in a hard, jagged line.
Everything runs in your browser. With WebGPU the graphics chip does the work in a few seconds; without it, the processor takes over and needs longer. Your photo is never sent anywhere, which matters for portraits, ID photos and product shots that are not public yet.
Tips for a clean cut-out
- Contrast helps. A dark subject in front of a light wall, or the other way round, gives the cleanest edge.
- Keep the subject in focus. Blurry parts come out half transparent, because the model cannot tell where they end.
- One main subject. In group photos or busy scenes the model may keep or drop people and objects you did not expect. Crop the photo first with the image cropper if it picks the wrong one.
- Check fine details. Loose hair, fur, glass and thin parts like chair legs are the hardest. Zoom into the preview before you use the result for print.
PNG, transparency and file size
With a transparent background the result is a PNG, because PNG keeps transparency and loses no quality. With a white or colored background you get a JPG at high quality instead. A cut-out PNG is often larger than the JPG you started with. For a website, convert it with PNG to WebP, which keeps the transparent background at a fraction of the size, or shrink it with the image compressor. To pull the words out of a photo or screenshot instead of the subject, use image to text.
Frequently asked questions
Are my photos uploaded?
No. The AI model runs inside your browser, on your own computer or phone. Your photos are never sent to a server, which also means there is no upload queue, no account and no limit per day.
Why does the first photo take longer?
The first time, your browser downloads the AI model and the engine that runs it, about 50 MB together. It keeps both, so on your next visit the tool starts right away and also works without an internet connection once the page is open.
How long does one photo take?
With WebGPU, which current Chrome, Edge and Safari versions offer on most computers and many phones, a photo takes a few seconds. Without it, the model runs on the processor and needs about 30 to 60 seconds per photo, depending on the device.
Which photos work best?
Photos with one clear subject: a person, a pet, a product, a car or a plant, in focus and standing out from the background. Fine hair, fur, glass and objects that blend into a background of the same color are the hardest parts, so check the edges in the preview before you use the result.
Can I put a different background behind the subject?
Yes. Choose White for product photos and online shops, or Color to pick any color. With a solid background you get a JPG, which is smaller and accepted everywhere. Keep Transparent for a PNG you can place on your own background in another app. Both are saved at the full size of your photo.
Is the background remover free for commercial use?
Yes. The tool is free, without a watermark, and the images you make are yours to use, also for a shop or client work. The AI model is ISNet by Xuebin Qin and colleagues, released under the Apache 2.0 license.