Background removal
The AI model cuts out the object and makes the background transparent or replaces it with a color. Photos stay on your computer. The pixel size does not change. Everything happens in the browser, files are not uploaded anywhere.
Background removal is the separation of the main object in a photograph from its surroundings. The result is used for product cards, avatars, documents, presentations and collages in which the object needs to be placed on another background. A transparent background is preserved in PNG and WebP formats.
A segmentation neural network is used to select the object. It analyzes the image, identifies the main object and creates a mask that separates it from the background. The mask is then applied as a transparency channel, after which the background pixels become invisible or are replaced with a chosen color.
Background removal from photos works right in the browser. The neural network model is downloaded once and kept in the cache, so on later launches the already downloaded version is used. A fast, smaller model and a more accurate one are available, the latter coping better with complex edges.
The result can be left with a transparent background, filled with white or with any other color. Additional cropping to the object removes the empty margins around it. For the JPG format, which does not support transparency, the background is always replaced with the chosen color.
The quality of the cutout depends on the image. The best results come from pictures in which the object clearly differs from its surroundings. Hair, fur, transparent objects and details of similar color may be selected less accurately. Very large photos are first reduced to 3000 pixels on the long side.
All calculations are performed on your device. Images are not sent to a server, so personal photographs stay confidential. Results are downloaded one by one or all at once. The page works on computers, tablets and smartphones.