Background Remover
Remove an image's background automatically, using an on-device AI model that runs entirely in your browser — your photo is never uploaded anywhere.
How does this remove backgrounds without a server?
This tool uses a machine learning segmentation model (Google's MediaPipe selfie segmenter, run through MediaPipe's on-device WebAssembly runtime) that identifies which pixels belong to the main subject and which belong to the background. The model and runtime files themselves — not your photo — are downloaded once from a public host the first time you use the tool, then run locally in your browser for every image after that.
Because the model works best on a single clear subject against a background, it's most reliable on portraits, product photos, and similar images — busy scenes with multiple overlapping subjects may give mixed results.
GuideHow to use it
Features & benefits
Why this runs entirely in your browser
Photos are often personal, so nothing here is uploaded to a server for processing. The only network activity is a one-time download of the public, open-source segmentation model itself (the same way a webpage downloads a font or script); your image data is never transmitted anywhere. Once loaded, everything runs locally in a WebAssembly runtime built into your browser.
Common uses
Cutting a product photo out for an e-commerce listing, isolating a portrait for a profile picture, or preparing a subject to composite onto a new background.
Frequently asked questions
A few things people usually want to know before trusting the numbers.
No. The image is processed entirely on your device in a local WebAssembly runtime. The only thing downloaded over the network is the AI model and runtime files themselves, once per browser session — never your photo.
The segmentation model (a few megabytes) is downloaded and initialized the first time you use the tool. After that, it stays in memory for the rest of the session, so subsequent images process much faster.
The model is tuned for a single clear subject — a person, product, or similar — against a background. Very busy scenes, multiple overlapping subjects, or very low-contrast images can make detection less reliable.
It needs a browser with WebAssembly support, which covers virtually all modern desktop and mobile browsers — no special hardware is required.
A PNG file, which supports transparency, so the cut-out subject can be placed on any background afterward.