The examples they give of how they “improved” the voices aren’t great. Version 1 sounded like actual people, while this version sounds like generic AI slop voice.