In the post-production of images, to achieve the effect of separating the portrait and the background, it requires time-consuming and laborious manual keying. The solid-color background keying can simplify this work, but it needs to limit the scene. AI portrait background keying, based on Meishe's self- developed machine learning, recognizes human silhouettes in images and videos, and separates them from the background.Combined with the rendering engine of Meishe SDK, it provides rich visual effects to enterprise users.