What if You Could Sing in Your Favorite Musician's Voice? | Holly Herndon | TED

TL;DR

In this content, Holly introduces Holly+, an AI trained on her own voice that can sing in multiple languages and explores the possibilities of using AI to create music and perform as different artists.

Transcript

(Overlapping voices sing) (Singing ends) (Applause) Hi, my name is Holly, and I’m an artist. That was my voice, but I didn’t sing that clip. I trained an AI on my own voice, and now she can sing anything in multiple languages. Her name is Holly+, and you just heard her perform “El Cant de la Sibilla,” which is a traditional song arranged by Maria A... Read More

Key Insights

• 🎶 Holly+, an AI developed by artist Holly, can sing in multiple languages and perform songs that Holly herself cannot. It uses timbre transfer to map one sound quality onto another.
• 🌍🎶 Holly+ can also perform Holly's own music, showcasing the potential for AI-generated vocals in the creative process.
• 🎵 Holly's concept of "spawning," where an AI can generate entirely new sounds based on a trained sound model, is similar to the musical tradition of sampling.
• 🔄 This raises questions about intellectual property and reanimating old media, necessitating new legal and conceptual frameworks.
• 🎭 Holly considers this technology as "identity play," allowing others to use her voice and intellectual property to create music.
• 💡 By making Holly+ freely available, she explores the potential for mutual benefit between artists and AI-generated music.
• 🎤 She asks the audience to consider who they would like to perform through and if they can imagine someone else performing as them.
• 🤝 Holly invites musician Pher to perform, showcasing the collaboration between human artists and AI-generated vocals.

Questions & Answers

Q: What is the name of the AI that Holly trained to sing in multiple languages?

The name of the AI that Holly trained to sing in multiple languages is Holly+.

Q: What is an example of a traditional song that Holly+'s AI vocalized?

An example of a traditional song that Holly+'s AI vocalized is "El Cant de la Sibilla," arranged by Maria Arnal in Catalan.

Q: How does Holly+'s AI perform timbre transfer?

Holly+'s AI performs timbre transfer by creating a machine-learning model of Holly's voice and mapping the timbre onto the performance of another sound.

Q: What is spawning, as described by Holly?

Spawning, as described by Holly, is a process where an AI generates an entirely new sound by using trained information about someone else's voice to perform as them.

Q: What does Holly hope to achieve by freely sharing versions of Holly+ online?

Holly hopes to achieve a mutually beneficial relationship with her AI voice by allowing people to use it while still maintaining the ability to approve certain derivative works.

Q: Who is the musician invited to perform onstage with Holly?

The musician invited to perform onstage with Holly is Pher.

Q: What song does Pher perform during the event?

Pher performs his own song called "Murky" during the event.

Q: How does Pher incorporate Holly+'s AI voice into his performance of "Murky"?

Pher incorporates Holly+'s AI voice into his performance of "Murky" by using a live version of Holly+ developed with Voctro Labs.

Summary & Key Takeaways

• The speaker, Holly, introduces her AI creation, Holly+, which is trained to sing in multiple languages using her own voice.

• Holly+ demonstrates its abilities by singing traditional songs in Catalan, German, and Spanish.

• The speaker discusses the concept of "spawning," where an AI can generate new sounds by combining the sonic properties of different sounds, and the potential ethical and legal implications of this technology.