Does this upload my photo anywhere?

No. The photo never leaves your browser tab. The AI model itself is downloaded from Hugging Face's CDN the first time you use the tool, but your image is only ever processed locally on your device.

Why does the first run take a while?

The background-removal AI model (MODNet) is about 26 MB and has to download before it can run. Your browser caches it after the first download, so future uses on the same device start almost instantly.

How fast is processing?

It depends entirely on your device's CPU or GPU — there's no server doing the work. A recent laptop or desktop typically finishes in a few seconds; older hardware or phones can take significantly longer. The tool shows which acceleration (GPU via WebGPU, or CPU via WebAssembly) your browser is using.

Which download format should I pick?

Use PNG or WebP if you need the transparent background preserved, for example to place the subject over another image. Use JPG if you want a flat image with a solid background color instead of transparency — pick the color before downloading.

Does it work on any photo?

It uses MODNet, a portrait-matting model trained specifically on people — it works best on photos with one or more people in them, including fine detail like hair. It isn't tuned for general product or object photography, so cutouts of things like furniture or food will be far less accurate.

AI BACKGROUND REMOVER

An on-device AI model cuts a person out of the background — works best on photos with one or more people in them. The model downloads once, then everything runs locally — no upload, no account.

+ Add a photo

Nothing uploaded — the image and the AI model both run in this browser tab.

How to remove a background

1
Add a photo — the AI model downloads once and is cached by your browser.
2
Click Remove background and watch the download and processing progress.
3
Compare the before and after with the slider.
4
Download the cutout as a transparent PNG, transparent WebP, or a JPG with a solid background color.

Questions

Does this upload my photo anywhere?: No. The photo never leaves your browser tab. The AI model itself is downloaded from Hugging Face's CDN the first time you use the tool, but your image is only ever processed locally on your device.
Why does the first run take a while?: The background-removal AI model (MODNet) is about 26 MB and has to download before it can run. Your browser caches it after the first download, so future uses on the same device start almost instantly.
How fast is processing?: It depends entirely on your device's CPU or GPU — there's no server doing the work. A recent laptop or desktop typically finishes in a few seconds; older hardware or phones can take significantly longer. The tool shows which acceleration (GPU via WebGPU, or CPU via WebAssembly) your browser is using.
Which download format should I pick?: Use PNG or WebP if you need the transparent background preserved, for example to place the subject over another image. Use JPG if you want a flat image with a solid background color instead of transparency — pick the color before downloading.
Does it work on any photo?: It uses MODNet, a portrait-matting model trained specifically on people — it works best on photos with one or more people in them, including fine detail like hair. It isn't tuned for general product or object photography, so cutouts of things like furniture or food will be far less accurate.