But will AI end concatenative synthesis supremacy?
It's already ended in TTS field, but in singing synthesis? Not anytime soon I think.
People still want to make their own voice available in singing synthesis programs, and concatenative synthesis remains popular (especially in free programs) and proven to be cost-effective.
As long as AI-based-voicebank-making is still considered hard, I don't think AI-based singing synthesis will go anywhere beyond professional/commercial scope.
edit: nvm, it does seem like Voiceroid really is discontinued.