A single meter reads one point and calls it the venue. We measure at every position that matters — then model the physics of everything in between.
Venues are not uniform. Two positions ten metres apart can differ by more than ten decibels at the same instant — and the corner that carries into the neighbours is rarely the corner with the meter on it.
Two speakers carrying the same signal reinforce in some places and cancel in others. Those patterns are fixed by geometry and wavelength — not noise, not error, and not visible to a single sensor.
Knowing a number was exceeded tells you nothing about where it came from, or whether pulling the mix down fixes it or simply ruins the room everywhere else.
Tuesday afternoon and Saturday night are not the same acoustic environment. One fixed limit is wrong for both — too strict for one, useless for the other.
This is live. The field below is computed in your browser right now with the same phasor summation that runs on the hub — drag the sources, change the frequency, move the weather. Everything responds because everything is derived, not drawn.
Drag either source to move it. Sensor markers show Δ — how far that position measures from what physics alone predicts.
Intensity falls with the square of distance. Correct when sources carry genuinely unrelated signals — and the point at which most systems stop.
Subs and PA cabinets fed from one bus are coherent. Summing their pressure as phasors is what produces the real peaks and nulls — adding power alone erases them.
Physics can’t see absorption, a hot amp or a shadowed corner. Each sensor’s residual δ pulls the prediction toward reality, weighted by distance — exact at the sensor, pure physics far from it.
The speed of sound is not a constant. It moves roughly 0.6 m/s per degree, which shifts every wavelength — and every interference null with it. A model pinned to 20°C is wrong for most of the British year, so we feed live conditions straight into the solve.
There is no training run and no model to redeploy. Each position keeps a running mean and variance for every hour of the day, updated on every reading, forever. Unusual is measured in standard deviations from that position’s own history — not against a number someone picked once.
A level spike at 4am could be a fault, a delivery lorry, or the cleaner’s radio. Knowing it happened is only half an answer — so alongside every level reading, each sensor also identifies what it is hearing, using YAMNet, Google’s audio event classifier trained on AudioSet. It runs on the sensor itself, every 30 seconds, across 521 recognised sound classes.
The classifier runs on the device, on a rolling buffer, and only its verdict is kept — a label and a confidence score. What crosses your network is levels, band energies, spectral descriptors and that label. Numbers, not recordings. There is no audio to store, leak or subpoena.
Individually calibrated measurement capsules at each position, reporting once a second to a hub that sits on your own network. It keeps monitoring, learning and alerting whether or not the internet is up.
Roughly four weeks building a real acoustic picture of your venue and learning its normal, while you keep full control of every decision. You see where the problem is, why it is happening, and whether the desk can actually fix it.
The limiter that follows is handed everything Phase 1 learned. Instead of one blunt cut applied everywhere, it makes small targeted corrections in real time, at the position the problem is actually occurring.
No obligation and no slides — a working install, live readings from every position, on a screen in front of you.
Book a walkthrough →