Harry Yeff (AKA Reeps One) is a ARS PRIX Electronica nominated artist and musician based in London. Yeff’s expertise in vocal musicianship and creative direction has generated an online global following, rendering over 100 million views and recognition as a pioneer of experimental vocalism.

---

Watch the complete keynote

---

Yeff’s path from musician to tech innovator was an unusual one. As an artist who specialised in creating sounds purely with his voice, he found that the more he pushed his voice, the more invitations he got to work with academics. Six years ago, he founded Reeps100, a studio to explore emerging voice tech.

Outside a musical context, you miss that people’s voices are not just unique, but impact on your sense of identity — and how people perceive you. The voice has an impact outside language. You’re all judging people by their voice, consciously or unconsciously.

If voice is part of the sense of self, could it be used to build another self? When he started sampling and creating voices, the idea of using AI to create a second self emerged. And he was pre-disposed to like the idea of a digital collaborator.

From chess to AI

Yeff was a teenaged championship chess player, and his Dad bought him a chess engine that would never get tired of playing against him. He ended up with what he describes as a “very warm” attitude to technology.

Game theory has the idea of the discomfort of opposition, that drives you to be better. Growth is in the discomfort. In chess, they say that every success dips you in brass: it makes you less adaptable and creative.

And he’s now finding that discomfort in creative directing projects beyond his life as a musician.

The AI second self

He’s created hundreds of hours of personally generated noise to train LLMs on. After spending years trying to make his voice behave like a machine, now he’s trying to make a machine create his voice. The result is an AI that sounds like him, but isn’t him.

That was the birth of Second Self, a project his did with MIT to produce an artwork interacting and compositing with a synthetic voice in an anechoic chamber. He created a duet, where he would sing phrases, and the AI would respond.

That collaboration opened up a string of research called Machines by Voice. And, after a while, the machines could create phrases he couldn’t keep up with — a discomfort he really didn’t like.

Making the embodied voice

Now, though, he’s working on visualising voices through Voice Gems, that create visual and physical embodiments of your voice. But you can apply that to synthetic voices, too… There’s a constant back and forth between his own skills and the synthetic voice that forces him to pull new things out of himself, like this project.

When they did an art installation in New York, which created a two-story-high visual representation of a voice, a small girl who was very shy, suddenly became loud and expansive when she saw the impact of her voice. This sense of the power of amplification is informing their new work: capturing multiple voices, and embodying the transitory nature of sound.

Why? They’re getting increasing request to recreate the voices of deceased people. We leave an audio footprint behind us, on our phones and devices. What can we do with them; what memorial can we build out of that?

The Voice Gems is an archive that allows us to capture moments of interaction. Children’s voices change fast – but they can capture Gems as they grow.

AI as creative mentor

His collaboration with the Leipzig Opera House was an exploration of mentorship. The project, critically, didn’t replace any human creatives. Instead, it used AI systems as a collaborator and mentor for the human performers.

People don’t embrace these emerging AI tools because they are afraid of their replacement; the loss of human creative. But there’s much more potential in the augmented human, who uses of AI as an assistant, collaborator, and mentor.

And that’s a narrative of hope.