AI Tools
These tools run a neural network locally using WebAssembly. The model is downloaded once and cached; your images are processed on your device and are never transmitted.
1 tools · all free · no sign-up
What “on-device” actually means here
Background removal normally works by uploading your photo to a server, running a segmentation model on a GPU, and sending back a cut-out. That is why most free tools want an account, and why their privacy policies are worth reading. The tools in this category do the same work in the browser tab instead, using WebAssembly.
The trade is a one-time download. The neural network weights are a real file — tens of megabytes — fetched the first time you use the tool and cached by the browser afterwards. On a fast connection that is a few seconds; on a slow one it is the slowest part of the whole operation. After that first fetch the tool works with no network at all.
There is also a no-download path. GrabCut is a classical segmentation algorithm rather than a neural network, so it ships with the page and starts instantly. It needs a hint from you about what is foreground and what is background, and in exchange it needs no model at all — which is often the better trade for a single image on a slow connection.