The voices are quite realistic but it fails in what says it can do:

> ... all while automatically matching what is being said with how it’s being said.

It doesn't do intonation correctly.