Really interesting, haven't listened to their output with high quality speakers or anything like that. Do poorly made human recordings have this problem or is this currently a signal of AI generation?