Ekheo: Giving birth to new voices
October 15, 2025
We sat down with experimental sound art duo Ekheo (Aude Langlois and Belinda Sykora) to talk about AI, identity, and creating music with their synthetic offspring
So how did you two meet, and what made you decide to work together?
Belinda: We met while studying sound studies at UdK Berlin in 2016. I came from performing arts—I was an actress, but always interested in abstract theater and the voice as this pure phenomenon. Aude has a more classical music background as a guitarist. We both were fascinated by the voice, just from totally different angles.
Do you remember the first time you worked together?
Belinda: I think, in our first project we were already Ekheo, working with recorded sounds, sensors and lights in a live performance.
Why “Ekheo”?
Aude: It comes from Echo, the Greek mythological figure. You know the story—her body disappears and she becomes just a voice. We were interested in this idea of voices being echoes of each other, but also the voice as an echo of something else, making other topics sonically perceptible.
How did you get from Greek mythology to AI?
Belinda: In this first project–which was not really thought through–we were already interested in this disembodied voice of Echo and what the relationships between bodies and voices entail.
Aude: And getting to AI was actually quite natural. When you work with voice, you’re already using microphones, processing… AI became just another way of manipulating and transforming the voice.
Belinda: Oh, and we also got really interested in voice assistants like Siri, especially around gender and identity.
Ah yes, the “I’d blush if I could” Publication by Unesco1…
Belinda: Exactly! Back in 2019-2020, if you said “you’re sexy” to Siri, that’s what “she” responded. We were asking: what’s the problem with gendered voice assistants? What would a gender-neutral voice even sound like?
That led you to create Leewa, your “digital child.” Can you tell us about that process?
Aude: This was during the Silicon Friend Camp2 in 2021. We used this AI oracle built on an early version of GPT-3 and basically interviewed it about our hypothetical child. We asked: “Belinda and Aude have a child—what’s their gender, their name, what do they like?”
And what did the oracle say?
Aude: It told us Leewa likes Walt Whitman, especially “The Body Electric.” Which is fascinating because Whitman is known for his gay writing, so there were all these unexpected layers that fed into Leewa’s identity.
Belinda: We then created a text-to-speech model using our voices, and those of other artists, to create this eerie collective voice of her. And then 3D-printed a puppet and performed with her. The goal was to create a sort of musical companion that can improvise with us.
You’ve also trained custom Neutone Morpho models with us using your voices. How did you approach building those datasets?
Aude: We created two very different datasets facing each other singing into microphones. The first was mathematical—we recorded chromatic scales, sustained notes, different vowels we sung together. Very structured, trying to blend our voices in this precise way.
Belinda: The second dataset was all the things an AI voice typically can’t do—breathing sounds, vocal textures, all those “body voices” that make human speech so complex.
Logic Session creating experimental vocal sounds 
Logic session creating vocal scales What’s it like performing with an AI version of yourselves?
Aude: It can be very weird. Imagine you’re singing and listening to yourself to adjust pitch and tone, but suddenly there’s this other voice that sounds like you but you have no control over it. You want to correct it, but you can’t because it’s not actually you.
That sounds like you’re giving away control…
Aude: In a way. We had to develop specific presets that we could recognize and switch between, so we know what sounds are coming. Most of our performances are 80% improvised, so we need to be able to react to each other—including Leewa.
You’re working on your first album now. How’s that different from your usual process?
Belinda: This is actually a big shift for us. For the first time, we’re not 100% improvising. We’re actually sitting in the studio composing songs with Leewa that people can listen to at home. It’ll be released in December.
What’s next for you and Leewa?
Aude: We’re exploring spatial audio and ambisonics. The interesting thing about synthetic voices is that audiences don’t have the same positional expectations as with human voices. We can place Leewa anywhere in 3D space and really experiment with that.
Belinda: We’re also applying for a four-year research grant in Austria. If we get it, we might finally build that physical puppet we originally envisioned for Leewa.
Speaking of the future—what do you think about AI’s role in music more broadly?
Aude: AI is a very good people pleaser. If you want to reproduce things that people already like massively, AI will nail it. But if you want to introduce new thoughts and push boundaries, you have to work with AI in a truly creative way.
Any concerns about AI replacing musicians?
Aude: Look, people were scared of microphones, they were scared of electric guitars. The key is ethics—we can’t be stealing from artists. But for creativity? There will be a lot of not so interesting AI music and a lot of amazing AI music, just like there’s always been boring music and great music.
One last question: what makes a voice a voice?
Belinda: That’s the big question we keep coming back to. Why do we say an AI voice isn’t a “real” voice? Maybe it is a voice. Maybe it’s the same conversation we have about intelligence. We really need to ask ourselves: what transforms a noise into a voice?
And have you found an answer?
Belinda: Not yet. But that’s what makes it interesting.
Ekheo will be performing on December 3rd at Silent Green in Berlin and their first studio album is about to come out at the end of this year. Follow their experiments in synthetic family dynamics at ekheo.com
Back