Bottom Line Up Front (BLUF): The tool removes unwanted objects from a photo right in your browser: paint over what should disappear, and the model redraws that area, continuing the background and the rest of the scene. The photo is never uploaded — the model download and all computing happen on your device. No tokens are charged and there is no paid API.
It cuts distracting details out of a shot. The input is a photo with an unwanted object; the output is the same frame without it: the model fills the painted area using its surroundings — sky, wall, grass, skin or pattern.
It is useful when you need to remove a passer-by or a car from the background, an accidental object, a wire in a landscape, clutter on a table or a glare on a bright wall. The tool also helps with editing: remove the extra first, then work with a clean background.
Upload a photo, pick a brush size and paint over what you want to remove: the mask layer shows clearly on top of the picture as a translucent colour. The “Clear mask” button resets the marking — you can paint again at any time.
Paint over the whole object and take some margin around it: the model redraws only what is visible through the mask. Several objects can be removed in one run — just paint them all.
The photo is fitted into a 512×512 square (black bars are added to keep proportions), the painted area becomes a binary mask, and the model redraws the hole step by step: twenty denoising steps produce a smooth, tidy result.
The result is blended with the original through the mask, so the edges of the redrawn area merge smoothly into the rest of the photo. The image is returned in its original proportions: the letterbox bars are cropped out and the download is a PNG.
The photo never leaves the browser: the model weights are downloaded once from an open storage and cached, then all computing happens locally. There is no paid API, so no tokens are charged.
The first run downloads about 1.27 GB of weights. It needs a browser with WebGPU support (a recent Chrome or Safari) and enough memory — roughly 2 GB. On weaker devices the tool falls back to the CPU and runs slower.
Nothing. The tool is free: no tokens are charged and there is no paid API — all computing happens in your browser on your device.
No. The file is never sent to a server: processing happens entirely in your browser, the photo never leaves your computer.
The first run downloads the model weights — about 1.27 GB. They stay in the browser cache, so later runs work without re-downloading.
Small and medium objects on a plain background: a table, sky, wall or grass. The more complex and larger the scene under the mask, the higher the chance the result looks blurred.
The model works at the fixed 512×512 resolution, so the photo is fitted into a square keeping its proportions. The result comes back in the original proportions: the bars are cropped out when you download.