Remove the background from an image

How it works

You don’t have to trace anything as with a manual background eraser: the cut-out is made by an AI model that decides, pixel by pixel, how much of the photo belongs to the subject. Best quality uses BEN2 (Background Erase Network), an open-source model that runs on your graphics card and holds on to the whole subject even when its colors are close to the background’s. Fast uses ORMBG (Open Remove Background Model), a smaller model built on IS-Net that runs on any device; its author trained it mainly on pictures of people, and it does best when the subject stands out clearly. Either way the mask is scaled to your full-size photo, so the result keeps every pixel of the original.

Getting a clean cut-out

• Contrast helps. A subject that stands apart from the background in color or brightness gives the cleanest edge.
• Keep to one subject per photo. Crop a group or a busy scene down to the part you want first with crop image.
• Use a decent-sized photo. Tiny images give soft edges; image upscaler can enlarge a small one first.
• Check the edges. Zoom in on hair, fingers and thin parts before you download. Shadows on the floor are sometimes kept as part of the subject.
• Fix what’s left with Touch up. The erase brush clears leftover background or a shadow, and the restore brush paints back a missing ear, strap or fingertip.

Transparent PNG, white or color

A transparent PNG keeps only the subject, ready to drop onto a slide, a design or another photo. Marketplaces such as Amazon ask for product photos on pure white (#FFFFFF), so choose White and JPG for those. A solid color suits profile pictures; JPG files are much smaller than PNG when there’s no transparency to keep.