So my complaint isn't that I dislike the voices - you did a really good job fine tuning what each character sounds like.
The issue is that AI voice generation is really bad at actual line delivery. The way they speak doesn't really sound like an actual person; they have a very start and stop delivery as if they had to double check the script mid-sentence; resulting in a brief pause.
AI voice generators also lack the ability to project emotion properly. It was especially noticeable for me when I got to the first erotic scene and Kairi's reaction is identical to how she would address anything else; no shock, confusion, or embarrassment over what the other version of her is doing.
That's the point when I backed out of the level to turn the voice volume off, since it was hurting my immersion.