Artificially generated but faithful to the original voice for people losing theirs or being unable to speak for various medical reasons. Obviously the fun/deceiving use-cases are much more numerous.