User input (such as what would move a player) is just a piece of data, such as a direction.
This sounds like an OpenCV question so I’d start with their forum support thread to understand the feature you want.
By “understand” I mean "integrate whatever face-watching feature you want, then analyze its output and generate either -1 (move left), 0 (no movement) or +1 (move right) output.
That is going to be the hard part, and that is ALL OpenCV.
Once you have the input, making the character move is trivial.