Background Remover

Because the actual neural network comes to you instead of your photo going to it. The segmentation model is tens of megabytes — that's what the upload sites run on their servers. Your browser caches it, so later uses start in seconds. On a slow connection, start it before you need it.

On typical subjects — a person, a product, a pet with a reasonably distinct background — results are comparable and the full-resolution output is free, which upload sites typically paywall. On the hardest cases (flyaway hair, transparent objects, subjects blending into busy scenes) dedicated cloud services with newer models can win. Try yours; it costs nothing.

Product photos on clean white for eBay, Mercari, and Amazon listings (cleaner photos measurably help sales), YouTube thumbnail cutouts, profile pictures, logos over any background, and compositing subjects into new scenes. The alpha channel means it drops onto any background without a halo.

The model needs working memory proportional to image size, and browsers cap what a tab can use. Photos beyond roughly 2500px on a side can exhaust it, especially on phones. Resize down first — our image resizer does it locally too — and note the segmentation quality doesn't meaningfully improve past ~2000px anyway.