Background Remover
Cut out people, products and pets automatically, keep the full resolution, and put them on transparency, a colour, a blur or your own picture. The AI runs on your device, so your photos are never uploaded.
Runs on your device. Nothing is uploaded.
AI model
Finds the main subject of any photo: products, pets, cars, people. Fast and small (4.6 MB). Switching keeps both results, so you can compare.
Loads with the first photo: about 19 MB once (WebAssembly build), then kept in this browser.
Edges
Removes the old background's colour from see-through edges such as hair, so they do not glow on a new background.
Background
Output
Saved as name-no-bg.png. Every file is drawn again from the pixels, so EXIF metadata such as GPS location is not copied.
No file at hand?Photo: U.S. Fish and Wildlife Service, public domain
How to use the Background Remover
- Drop one or more photos or a whole folder, paste an image with Ctrl+V, or click to choose files. The first photo downloads the AI runtime and model once (about 19 to 31 MB depending on your browser, plus 26 MB if you use the People model); after that they start from your browser's storage.
- The background is removed automatically, one photo after another. Pick the General model for products, pets, cars and most photos, or People for portraits where fine hair matters. Switching keeps both results.
- Drag across the preview to compare the original and the result. Under Edges, keep refinement on Standard or choose Fine detail for hair and fur; use Shift edge to trim a thin halo and Feather to soften the edge. Decontaminate colours removes the old background's tint from see-through edges.
- If the AI missed something, choose Keep or Remove and paint over it. Set the brush size and softness, zoom with Ctrl+wheel or a pinch, hold Space to pan, and undo or redo any stroke. Touch-ups are applied to the full-resolution mask.
- Under Background, keep it transparent or put the subject on a colour, a blurred copy of the original, or a picture of your own. Under Output, choose PNG or WebP (both keep transparency) or JPG, crop to the subject with padding or as a square, and set the size.
- Download the current photo, or all of them as a ZIP. Files are named name-no-bg and carry no EXIF metadata.
What it does well
Full resolution with clean edges. Many in-browser removers return the small mask the model works on, scaled up and blurry. Here the mask is rebuilt at the size of your photo with an edge-aware guided filter, so hair, fur and product edges stay crisp where they are crisp and soft where they are soft. Shift edge, Feather and colour decontamination finish the edge for the background you choose.
You can fix what the AI gets wrong. Keep and Remove brushes with size and softness paint on the full-resolution mask, with zoom, pan, undo and redo, and a before/after slider shows exactly what changed. Every photo in a batch keeps its own touch-ups.
Private and ready to use. Two openly licensed models run on your graphics card or processor, nothing is uploaded, and the files are saved as PNG or WebP with transparency or as JPG, cropped to the subject if you like, at the size you need, one by one or as a ZIP.
Frequently asked questions 8
Are my photos uploaded?
No. The AI model runs in your browser, on your graphics card through WebGPU where available or on your processor through WebAssembly. The page downloads the model and its runtime from this site; your photos, the masks and your touch-ups never leave your device. A background picture you add stays in memory only. Your usage history, if you keep one, records only settings such as the model, format and background type, never images or file names.
Which AI models does it use, and what is their licence?
General uses U²-Net small (u2netp) by Qin et al., released under the Apache License 2.0. People uses MODNet by Ke et al., a portrait matting model whose code and models are released under the Apache License 2.0. Both run with ONNX Runtime Web (MIT licence). All files are served from this site and checked with a SHA-256 hash before use.
How big is the first download, and is it stored?
About 31 MB with WebGPU (a 27 MB runtime plus the 4.6 MB General model) or about 19 MB without it; the People model adds 26 MB when you first choose it. The files are kept in your browser's storage, so later visits start without downloading. The AI model section shows how much is stored and has a button to delete it.
How fast is it?
With WebGPU (current Chrome and Edge, and other browsers that support it) a photo usually takes well under a second once the model is warm; the very first photo also prepares the graphics card, which can take a few seconds. Without WebGPU the General model runs on the processor and takes a few seconds per photo. The People model always runs on the processor (its results on the graphics-card path are not reliable yet) and takes about 5 to 10 seconds per photo on a typical laptop. Saving a large photo at full resolution adds a moment, because edge refinement and colour clean-up run on every pixel.
Will the result keep full resolution?
Yes. The model works on a small copy of the photo, then the mask is scaled back up and refined with a guided filter that follows the edges of the full-resolution original, so the saved file has the same pixel size as your photo unless you choose another size.
What does it handle badly?
Transparent and reflective objects (glass, bottles, water), fine structures with background showing through (fences, nets, bicycle spokes, wire), subjects that blend into a background of the same colour, very low-contrast or dark photos, and pictures with several equally important subjects, where it may keep or drop the wrong one. The People model is only meant for people. Use the Keep and Remove brushes to fix such spots.
Which formats can I open and save?
It opens JPG, PNG and WebP, plus GIF (first frame), BMP and AVIF where your browser can decode them. It saves PNG or WebP with transparency, or JPG on a colour of your choice. If your browser cannot write WebP (older Safari), PNG is saved instead.
Is there a watermark or a limit?
No watermark, no sign-up and no limit on the number of photos. Large batches simply take longer, because every photo is processed on your device.
More free tools
- Object & Blemish Eraser Paint over people, objects, text, wires or spots and erase them with AI that runs in your browser. Spot heal for blemishes. Full resolution, no upload.
- Face Blur & Pixelate Find faces automatically, even small ones in group photos, and blur, pixelate or cover them. Draw over plates or badges. Batch, ZIP, no upload.
- Screenshot Stitcher Join scrolling screenshots into one full-page image. Overlaps and sticky headers are removed automatically. PNG, JPG, WebP or PDF.