CCM Multi-Mic Recorder

Conduct of Chapter Meetings · one line per officer · webcam sync
Parliamentary Owl Consulting · parliowl.org

Microphones

Each DJI Mic 3 receiver takes up to four transmitters. For seven officers use two receivers on USB, both in Q (Quadraphonic) mode: TX1–TX4 on the first, TX1–TX3 on the second. Every transmitter is saved as its own mono file.

Input channels and the officer on each
DeviceChannelWho is on itLevel

Tip: to tell the two receivers apart, tap one transmitter's windscreen and watch which row moves.

Sound check

  1. Everyone seated, transmitters clipped on, receivers on Q mode.
  2. Press Start sound check. Each officer in turn reads their opening ceremony line at full debate volume (about 5 seconds each). One member then says "Second" from their seat, normally.
  3. Press Finish. Make the gain changes shown on the receiver for that transmitter, then run it again until every line is green.

Camera & seats

Aim the webcam at the members' side of the table so every seat is in view. Then: choose an officer and click on their head; repeat for each seat. Drag the gold line to just above the seated heads. With everyone seated, press Everyone is seated. When someone rises, their head crosses into the band above the line.

Live webcam view of the members' seats

Record

00:00

Lines

Now speaking (backup marks; keys 1–7)

Recordings

Analyze lines up the mics, works out who spoke when (timeline.csv), makes one cleaned 16 kHz copy per officer for Whisper, and a review video if ffmpeg is installed. Transcribe sends only the speech pieces of each officer's copy to Whisper large-v3-turbo on Cloudflare, so the transcript comes back labeled by officer.

Saved recordings, newest first
RecordingLengthLinesStatusActions

Settings

Recording

Transcription (Cloudflare Workers AI)

Uses the same model as the CCM app, @cf/openai/whisper-large-v3-turbo. The token stays in config.json on this laptop. Make a token with the Workers AI permission only.

DJI Mic 3 settings for CCM

Receiver and transmitter settings and why
SettingUseWhy
Recording mode (receiver)Q – QuadraphonicEach transmitter on its own channel over USB. In S mode with 3–4 transmitters, TX1+TX3 and TX2+TX4 are mixed together.
Noise cancellingOff (Basic only if HVAC is loud)Gothenburg showed gating cuts quiet words and makes the mic look dead between words.
Adaptive gain controlOffGain that rides up and down hides who is talking; speaker detection compares levels between mics.
Voice toneRegularFlat sound transcribes best.
Lossless audioOn48 kHz 24-bit over the air; range is not an issue in a contest room.
Transmitter gainSet in Sound checkTarget: speech around −24 dBFS, peaks near −12. Leaves room for a raised voice.
Internal recording (each TX)On32-bit float backup on the transmitter itself, in case of a wireless dropout.
Auto power-offOffA mic that sleeps during a long secretary's report is a lost line.

On the laptop

  • Windows: Settings → System → Sound → each DJI receiver and the webcam → turn Audio enhancements off; set format to 48000 Hz. This app uses WASAPI, which shows all four Q channels (as Audacity does).
  • Mac: Audio MIDI Setup → each receiver should show 4 input channels at 48 kHz. Allow the Terminal/app to use the microphone and camera.
  • Plug both receivers directly into the laptop (not a passive hub). Plug in the charger: two receivers and a webcam are a real load.
  • Webcam mic: it records the room as a reference and timing check. It never decides who is speaking.

On the members

  • Clip the transmitter at the center of the chest, about a hand's width below the chin, outside clothing. Keep the windscreen on to stop breath pops.
  • Keep hair, lanyards and the official jacket collar off the mic; FFA jacket zippers and corduroy rub loudly.
  • Members seated close together will show on each other's mics. That is fine; the app turns each officer's neighbors down. The sound check warns only when the spill is strong enough to confuse it.