Its important for the camera that it only does one thing at a time. So in code you must break out of all Open-CV stuff, then do the screen/video recording, then return back to Open-CV if so desired. This fellow figured it out for his bird capturing Raspberry Pi here - Object and Animal Recognition With Raspberry Pi and OpenCV - #117 by Roger196291
That system works by running object detection, searching for birds. Then when one is identified it stops all Open-CV tasks, takes a full HD photo, then reverts to object detection for more birds. If you mimick what he has done but for the Face Identification your system should be up and running perfectly 
