How we transform meetings with AI-driven audio

Taking our first bold step into audio with Huddly C1 was a big deal. Luckily, our team was ready for the challenge. Building on years of experience with AI-driven video, they applied many of the same principles to sound.

“The basic techniques are the same,” says Robin Andlauer, who’s been with Huddly for almost four years: "The training, the tooling, the way we test things. There are many parallels between computer vision and audio modeling, but audio brings its own challenges. What matters most is that it sounds natural, not synthetic.”

The step change is magnitudes. It’s a huge leap forward in every sense.– Danny Ward, BT Group

Our users’ feedback speaks for itself. Take BT Group, who have already deployed the C1 at their Manchester flagship office and are rolling it out to additional locations: “The step change is magnitudes. It’s a huge leap forward in every sense. You can really tell the difference when you move from a room with a legacy bar to one with the C1. Everything seems brighter, clearer, and the audio is great,” says Danny Ward, Design and Strategy Lead for AV.

In this conversation with Robin, we pull back the curtain to look at how C1 uses AI to transform meeting room audio.

The challenge of meeting room audio

Great communication starts with clarity, and in a meeting room, clarity depends on how sound travels from the person speaking to the microphone. “When the microphone is close, your voice comes through clean and clear,” Robin says. “But when it’s a few meters away, the sound bounces off the walls and comes back with a delay that blurs the sound. We call that reverberation, or reverb.”

This makes voices harder to follow and increases cognitive load — the effort it takes for your brain to separate speech from noise. Modern meeting rooms make this even more difficult. Glass walls, hard floors and minimalist design reflect sound in unpredictable ways, while other real-world factors like room shape, voice type and background noise further complicate how audio behaves.

Clear natural voices, driven by AI

AI-driven audio adapts to these real-world conditions, letting us go beyond what traditional signal processing can do. How? C1’s AI is trained on real scenarios to recognize how natural voices sound and separate them from unwanted echo, noise and reflections. Powered by neural networks, it applies those learnings in real time to reduce echo, suppress noise and enhance clarity.

This intelligence works hand in hand with C1’s precision hardware. “The microphones are already very good,” Robin says. “The AI just helps them perform at their best when the acoustics aren’t.” Together, sixteen microphones capture every detail, while the AI refines what you hear, keeping voices clear and natural across different spaces and meetings.

“The idea of AI can sound scary,” Robin admits, “but we don’t use it to create or change anything. We just make what’s already there better and easier to hear.”

Built for what comes next

At Huddly, we’re lucky to have a team that’s both talented and unafraid to think outside the box.

“The audio team brings together people from machine learning, acoustics, signal processing, and more,” says Robin. “That’s a strength. It makes us open to trying completely new approaches.”

And with twenty times more AI processing power, C1 gives the team the headroom to keep reimagining what meeting room audio can do.

Ultimately, the goal is for it to sound like everyone has a microphone right in front of them, so voices come through clear and natural no matter where they are in the room.– Robin Andlauer

“We’re constantly refining the models. There’s always more to learn. Ultimately, the goal is for it to sound like everyone has a microphone right in front of them, no matter where they are in the room.”

To experience the C1 difference, book a demo today.