Design · 3 min read
How AI Background Removal Works
Cutting a subject out of a photo used to mean careful work with selection tools. AI models now do it in seconds — and some run entirely in your browser, so the photo never leaves your device.
Try it yourself with the related tool.
Remove a background →Advertisement
Segmentation in a nutshell
A segmentation model is trained on many images with hand-made masks. For each pixel, it predicts how likely it is to belong to the foreground. The result is an alpha mask — opaque for the subject, transparent for the background, and partly transparent along soft edges like hair.
Photos that work best
A clear subject, good lighting, and contrast with the background give the cleanest edges. Fine hair, glass, smoke, and subjects that match the background colour are hardest. If the edge is rough, try a photo with more separation from the background.
Using the result
Save as PNG or WebP to keep transparency — JPEG has no alpha channel and will fill the background with white or black. A soft shadow under the subject helps it sit naturally on a new background.