offer some kind of distortion to avoid exact replication

This is just an idea. Some folks might have privacy concerns about sharing clear voice recordings, as they can be used to synthetize and create TTS models of other people without authorization. Maybe offering some kind of optional distortion would be a good feature to add.