Background removal (also called cutout or matting) is the process of separating the subject of an image — a person, product, pet, or logo — from everything behind it, and outputting the subject alone on a transparent background. What used to take careful pen-tool work in an editor now takes one click: an AI model reads the photo, decides what the subject is, and returns a PNG with an alpha channel.
How it works
Under the hood this is a segmentation problem. The model is trained on millions of images paired with hand-made masks, so it learns which pixels belong to the foreground and which to the background. Modern matting models go beyond hard masks: they estimate a soft alpha value for every pixel, which is what lets them keep individual hair strands, the translucency of a glass, and the faint edge of a drop shadow instead of cutting a jagged outline around them. The result keeps the original pixel dimensions — nothing is resized or cropped — so the cutout drops straight into a design.
What it's best at
- Product photos — clean cutouts for store listings, comparison grids, and ads that need a uniform background.
- Portraits — profile pictures, ID photos, and team pages with a consistent backdrop.
- Logos and handwriting — scans and screenshots that need a transparent edge before they go on a page.
- Compositing — the first step before placing a subject into a new scene or an AI-generated background.
What a good input includes
Give the model a clear subject and some contrast with the background: a product on a plain surface or a person against a wall cuts out better than a busy street scene with three overlapping people. Even lighting helps; harsh backlight blends hair into the background. Use the highest-resolution copy you have — edges are estimated from the pixels available, and a crisp source gives a crisp cutout.