Is it possible to make a “smart” 3D model using three.js?

Question

JavaScript

Hello.
I need to solve the following problem!
I have a window that displays a 3D model provided to me in .dae format and run with Collada. Its background is a stream from my webcam.
Implemented the ability to determine the user's face in this stream. Is it possible to somehow influence the model when determining a face? That is, for example, the user's face was determined in certain coordinates of the window, I need my model (I have an elephant) to turn its head and look at the coordinate area where my head is located. Or he waved his trunk at that area of ​​​​coordinates, etc.
I would be grateful for any answers, as I myself am new to this business.