Background Remover

No. The AI model is downloaded to your browser and runs there; your photos are never sent to Toolkitta or anyone else. Keep the tab open until you download or send your results — reloading clears the batch. Website hosting and performance measurement are described in our Privacy notice.

The background model is about 46.55 MB. The first run downloads it from Toolkitta and stores it in this browser, so later visits start in a few seconds. Clearing site data removes it, and a private window downloads it again each time.

Clear subjects — people, pets, products, cars, food — usually come out clean. Fine hair, fur, glass, smoke and subjects that blend into the background can lose or keep bits, so open the preview and compare with the original before you use a result. There is no manual eraser yet.

Yes. The model decides what is subject at 1024 × 1024, and that mask is then stretched smoothly over your full-size photo, so every original pixel of the subject is kept — up to 40 megapixels. Nothing is watermarked.

Choose White (or Colour for any other shade) before you run. JPG is then offered as well and is much smaller than PNG for photos. Transparent results can only be saved as PNG or WebP, because JPG has no transparency.

It crops the result to the subject with a small margin, which is handy for stickers, logos and product shots that you will place elsewhere. Leave it off to keep the original picture size.

Browsers with WebGPU (recent Chrome and Edge) run the model on the graphics card in about a second. Other browsers use the processor, which can take ten seconds or more per photo. Large batches on phones may run out of memory — do a few at a time.

No. Every result is rebuilt from pixels, so camera settings, GPS location and other embedded metadata are removed. Orientation is applied first, so photos are not rotated.