FaceID doesn’t use the standard visual camera.

It’s (for example) similar to the Kinect. It projects a pattern of structured light in IR then looks at that with an IR camera to be able to determine depth.