Audio Record

Record Signal (or other) video calls on Linux with both system audio and microphone, using PulseAudio echo cancellation to prevent feedback.

Canonical scripts:

• ~/bin/record-call.sh
• ~/bin/signal-record.sh

This repo keeps notes, plans, and supporting material. The runnable signal-record.sh now lives only in ~/bin/.

Requirements

• Linux with PulseAudio
• ffmpeg
• PulseAudio echo cancellation module (module-echo-cancel)

Configuration

Edit record-call.sh and update these variables to match your hardware:

MONITOR="alsa_output.pci-0000_c6_00.6.analog-stereo.monitor"
MIC="alsa_input.usb-EMEET_HD_Webcam_eMeet_C950_A230803002402311-02.analog-stereo"

To find your device names:

# List audio sources (microphones and monitors)
pactl list short sources

# List audio sinks (speakers/headphones)
pactl list short sinks

Usage

# Record with default name (call-TIMESTAMP.m4a)
./record-call.sh

# Record with custom name (meeting-TIMESTAMP.m4a)
./record-call.sh meeting

Press q or Ctrl+C to stop recording.

How It Works

The script:

1. Loads PulseAudio's WebRTC echo cancellation module
2. Captures system audio (what you hear) via the monitor source
3. Captures microphone input via the echo-cancelled source
4. Mixes both into a single AAC audio file

Experiments

The experiments/ folder contains alternative approaches tried for echo reduction:

Script	Approach
v1-volume	Reduce mic volume
v2-amerge	Use amerge instead of amix
v3-highpass	Highpass filter at 80Hz
v4-echocancel	PulseAudio WebRTC echo cancellation ✓
v5-gate	Noise gate on mic
v6-combo	Combined approach

See experiments/ECHO-EXPERIMENTS.md for details.