@VooDooMedic @HonkHonkBoom @charliebrownau I haven't had this issue before with other AI voices. This particular AI voice was trained on data that was exaggerated to sound sexy, which happens have strong S's. So when it was recreated, it over-emphasized the high frequency S's. This matches up with @HonkHonkBoom's graph which peaks around 7-8 kHz.