TED Talks Daily What if you could sing in your favorite musician's voice? | Holly Herndon
Sep 7, 2022
What if you could sing like your favorite artist? Holly Herndon introduces a groundbreaking AI instrument called Holly+ that allows users to mimic her voice. A live demonstration showcases this technology, revealing new dimensions in music creation. The discussion dives into 'spawning,' an AI technique creating unique sounds from existing vocal recordings, raising fascinating questions about identity and intellectual property. Finally, the episode beautifully captures the emotional connection of live music, blending human and digital vocals.
AI Snips
Chapters
Transcript
Episode notes
Holly+ Demonstrations
- Holly Herndon trained an AI on her voice, creating Holly+.
- Holly+ performed a Catalan lullaby and a German excerpt, showcasing its multilingual capabilities.
Timbre Transfer
- Holly+ uses timbre transfer to map the sound quality of Herndon's voice onto other performances.
- This involves creating a machine learning model trained on Herndon's diverse vocal range.
Spawning vs. Sampling
- Herndon introduces "spawning," generating new sounds from trained data, like a 21st-century sampling.
- Spawning allows performing as someone else based on learned information, unlike traditional sampling.