TonyS I am not sure how MS Photos works, I failed to make it work for my own photo archive. It just randomly ignored part of my photos, so I just gave up.
The main issue with indexing people in videos is processing time. Now, with running face detection just once per image, it already takes significant amount of time to process large archive. With videos, face detection (and then face recognition for each detected face) should be called multiple times, ideally several times per each second of the video. Doing that on average personal PC may take forever for large collection of videos. So this is definitely doable, but in practice this is not only software issue, this also may require additional dedicated hardware for processing.