A mind-boggling application of machine learning, created by Holly Herndon and her collaborators, that employs timbre transfer to map a singer's voice signature onto the performance of another—including in realtime, as demonstrated in the linked TED Talk.
