Minut technology

AudioID

Patented privacy-first sound recognition

AudioID is Minut's proprietary machine learning technology that identifies real disturbances and critical sounds. Engineered for absolute privacy, it processes sound patterns directly on the device, ensuring conversations are never recorded or recognized, and raw audio never leaves the room.

Illustration of an on-device sound signature: frequency over time, with an alarm recognized in a living room

50+

sound classes AudioID is trained to recognize

Breaking glass

Inside the R&D

Engineered to understand sound, not just measure it

Traditional decibel monitoring measures how loud a space is, but cannot determine what made the sound.

That’s why Minut’s team of engineers trained AudioID on over 50 sound classes, so our sensors can separate real disturbances, like a loud party or a safety alarm, from the ambient noise of everyday life.

Sound recognition

Precision sound recognition

Using convolutional neural networks, AudioID identifies the signatures of standard household safety alarms, like fire and CO alarms, and breaking glass, in real time.

Trained on thousands of data points, including look-alike sounds, it’s precise enough to tell breaking glass from a dishwasher being emptied.

Living room

Safety alarm pattern detected

Alarm recognized

Kitchen

Breaking glass detected

Glass break

Kitchen

Dishwasher being emptied

Ignored

Minut sensor mounted outdoors next to trees moving in the wind

Sound suppression

Fewer false alerts, by design

Standard noise monitors measure loudness alone, so a windy night or a vacuum cleaner can register as a disturbance or party. AudioID is trained to recognize the specific sounds that matter.

Engineered for the elements.

AudioID is trained to tune out environmental and weather noise, delivering reliable noise alerts inside, and out.

Edge processing

Audio never leaves the sensor

Sounds are processed in real time on the sensor and immediately discarded. Only privacy-safe data is sent to the cloud: never voices, never recordings.

Minut sensor

On device

Detect

The microphone picks up a sound in the room.

Transform

AudioID turns it into an anonymous sound signature on the sensor.

Discard

The raw audio is deleted immediately. It is never stored or sent.

Minut cloud

Receives only event labels and noise levels.

Minut user