The system is called simultaneous two-factor verification of identity (C2FIV). It requires the user to record a short video of the face. You also need to read a specific phrase aloud. Then you need to download the video to the device for authorization. The system will further require both the face itself and a certain gesture to be raised.
C2FIV reportedly relies on a neural network that has been trained on a dataset of 8000 videos. The people on the rollers winked, smiled, or raised their eyebrows.
The system should be more reliable. As its creator says, even if someone gains access to your device, they will never know which gesture you use to unlock.