jtruk, I'm not sure whether to pursue this...
My friend Nick died last year. His wife is looking to gather together a folder of videos of him because she and the girls miss his voice.
I'm wondering whether an AI system can make a model from this mixed corpus of speech and reproduce it to an authentic standard, so that 'he' can do things like wishing the girls a happy birthday.
Would this be (1) feasible (2) healthy?
(Let's assume I've watched every Black Mirror twice, so I don't need pointing there!)