A thud means nothing until you know where it happened

We didn't set out to build an audio product.

Eventag started in a neuroscience lab at the University of Oldenburg. The researchers there record EEG outside the lab, on people walking down streets, sitting in offices, talking to friends. The brain data comes back fine. What's missing is what the person was hearing at the time.

In a lab that part is easy. A computer plays a sound and writes a marker into the recording at the same millisecond. Outside, nobody controls the sounds. So researchers record video, take notes, and then spend days going back through the footage to mark when a car passed or someone started talking.

We wanted to automate that. Sentien is what came out of it.

Labels aren't enough

The obvious approach is a sound classifier. You give it a short clip and it tells you what the sound is: dog bark, siren, door slam. The field is called acoustic event detection, and it has been around for years.

It didn't answer the researchers' question. A loud thud and raised voices mean one thing in a busy restaurant at lunchtime. They mean something else in a hospital corridor at 3am. The label is the same in both places. What it tells you isn't.

To make sense of a sound you need to know where it happened and what was going on around it.

What Sentien does instead

Sentien takes audio, recorded or live, and returns four things for each moment:

• What the sound was, with a confidence score
• What kind of place it was, like an office, a busy street or a quiet room
• What was probably going on, like people working together, someone moving around, or quiet
• How the sound changed over time

All four come back together, timestamped, as JSON or CSV. You get structured data you can work with, and nobody has to listen back through the audio.

For the researchers, that meant lining up brain signals with what was happening around the person, without days of manual work.

Why we think this goes beyond research

Once it worked, we kept noticing how many devices have a microphone and no idea what's happening around them.

A smartwatch sees your heart rate jump to 120. It can't tell whether you're running for a bus or sitting in a tense meeting. A care home wants to know if a resident is alright at night but can't put cameras in bedrooms. A building wants to know which rooms are actually being used.

None of these need a transcript of what anyone said. They need to know what situation the device is in. We think that's the gap Sentien fills, and we're opening access now to find out how right we are.

Where we are

Sentien is in early access. If you're building something that would work better if it understood its surroundings, join the waitlist and tell us what you'd use it for. Those answers shape what we build next.