Capture a room

Capture a room

Banter can measure the room you are standing in — playing a test sweep through a speaker, recording it back through a microphone, and computing the room's acoustic fingerprint (its impulse response). That fingerprint feeds the room simulator so acoustic-comms tests can be replayed against your real room, and it gives you two plain numbers describing how live or dead the space is.

This guide shows how to run a capture from the node UI.

Before you start

• A speaker (audio output) and a microphone (audio input) reachable by the app. They do not need to be the same device. A USB audio interface works well.
• A quiet room. The measurement is a swept tone that rises in pitch over a couple of seconds; background noise during that window degrades the result.
• Set a comfortable output volume. Too quiet and the sweep is buried in noise; too loud and the speaker distorts. Aim for clearly audible but not straining.

Open the Room Capture screen

Reach it either way:

• The Capture Room button on the top toolbar, or
• Utilities → Capture Room in the hub menu.

Run a capture

1. Output device — pick the speaker to play the sweep through.
2. Input device — pick the microphone to record with.
3. Sweep settings (the defaults are sensible; adjust only if needed):
   • Start / End frequency — the pitch range the sweep covers. The default spans the audible band.
   • Duration — how long each sweep lasts. Longer sweeps push more energy into the room and improve the result in a noisy space.
   • Tail — how long to keep recording after the sweep ends, so the room's reverberation decays into the recording rather than being cut off.
   • Repeats — run the sounding several times and combine the results for a more robust measurement. Vary length (on by default) changes the sweep length between repeats so time-varying noise averages out.
4. Press Capture. The progress bar tracks play-and-record; keep the room quiet until it finishes.
5. When it completes, the results show the room's:
   • RT60 — reverberation time in seconds: how long sound takes to decay by 60 dB. Small treated rooms are ~0.2–0.4 s; a bare hall can exceed 1 s.
   • DRR — direct-to-reverberant ratio in dB: how much of the energy arrives straight from the speaker versus bouncing around. Higher = drier / closer.

If it fails, the status line explains why (most often the wrong input device, a muted mic, or a device that vanished). Fix it and press Capture again.

Save the result

Press Save. This writes two files:

• a .wav holding the measured impulse response (a float-WAV that the room simulator's measured-RIR engine and the convolution unit both load directly), and
• a .json scene calibrated to the measured RT60, so the geometric room model can reproduce a room of the same liveness.

Keep captures per speaker/mic placement — moving the devices changes the fingerprint, because the measurement bakes in the speaker, microphone, and their frequency responses along with the room itself.

What to do with it

A saved capture is the ground-truth anchor for acoustic-comms work: the room simulator can replay a codec through your room (measured engine) and cross-check it against the parametric geometric model, tightening both. See Room simulation in Banter for how the two models work together.