You've probably been on this call. You're at a café or a metro station. Your earbuds have noise cancellation, so the world has gone quiet for you. Then the person on the other end says, "Sorry, can you repeat that? There's some background noise."

That's the gap between active noise cancellation (ANC) and voice isolation. They sound similar, and most spec sheets bundle them together, but they solve opposite problems.

Active noise cancellation protects your ears. Microphones on the outside of the earbuds listen to the world around you, generate the inverse of that sound wave, and play it into your ear. The two cancel each other out. The result is quiet. You hear your music, your podcast, your call, without the world leaking in.

But notice the direction. ANC works on the path from the world to you. It does nothing for the path from you to the world. The noise around you is still there, and your microphone is still picking it up and sending it straight to whoever's listening, whether that's a colleague on a call or an AI model trying to transcribe what you said.

ANC cancels the noise around you. It doesn't stop that noise from reaching the receiver.

Here comes voice isolation. It protects your voice. Instead of silencing the world for you, it separates you from the world before your voice leaves the device.

This is a harder problem. It means using multiple microphones to find out where the sound is coming from, focusing on the direction of your mouth, and suppressing everything else. Done well, the person on the other end hears you as if you were in a quiet room, even when you're not.

It also means you don't have to raise your voice to be understood. and that matters more than it sounds.

Why this matters for voice AI?

We think voice is the most natural interface humans have. Speaking is faster than typing, and it's how we've always thought out loud. But most people still don't talk to their devices in public, and it isn't because the models aren't good enough. It's because speaking loudly to a machine in a crowded place feels strange. You repeat yourself. The transcription comes out wrong anyway because the room got into the mic.

Voice isolation removes that friction. When your device can hear you clearly at a normal, even soft, speaking volume, voice stops being something you save for when you're alone. It becomes something you can use anywhere.

project mnemosyne does both. We built the earbuds with ANC and voice isolation because you need both directions to work.

ANC so you can hear clearly, wherever you are. Voice isolation, so you can be heard clearly, wherever you are, by the people you talk to and by the AI/agents you work with.

Make more of your voice. Use your voice everywhere.