| name | test-input-level |
| description | Record a short sample from the active input device, compute peak/RMS levels, and recommend gain adjustments. Use to verify a mic preset is producing clean, well-leveled audio for voice dictation or recording. |
test-input-level
Inputs
duration_sec — recording length. Default 5.source_name — optional PipeWire/Pulse source name. If absent,
records from the default source.
Steps
1. Pick a tool
Prefer parecord (pulseaudio/pipewire-pulse) since it's near-universal
on modern Linux. Fall back to pw-record. Confirm one is installed:
command -v parecord >/dev/null || command -v pw-record >/dev/null \
|| { echo "Need parecord or pw-record. Install pulseaudio-utils or pipewire-cli."; exit 1; }
command -v sox >/dev/null \
|| { echo "Need sox for level analysis. Install with your package manager."; exit 1; }
2. Record
TMP=$(mktemp --suffix=.wav)
trap 'rm -f "$TMP"' EXIT
if [ -n "$source_name" ]; then
parecord --device="$source_name" --file-format=wav "$TMP" &
else
parecord --file-format=wav "$TMP" &
fi
REC_PID=$!
sleep "${duration_sec:-5}"
kill "$REC_PID" 2>/dev/null
wait "$REC_PID" 2>/dev/null || true
Tell the user when recording starts and stops — they need to be
talking at a normal level during the window.
3. Analyze
sox "$TMP" -n stats 2>&1
sox stats reports peak (Pk lev dB), RMS (RMS lev dB),
DC offset, crest factor, etc.
4. Recommend
Apply these heuristics:
- Peak above -3 dB → too hot. Reduce input gain or autogain target.
- Peak below -20 dB → too quiet. Raise gain.
- Sweet spot for voice dictation: peak between -12 and -6 dB,
RMS around -20 to -16 dB.
- DC offset > 0.01 → check that DC removal / high-pass is on in
the active preset.
- Crest factor > 20 dB → very dynamic; compression / autogain
may help dictation accuracy.
Report the numbers and the recommendation. Offer to:
- Re-run after the user adjusts gain in the OS mixer or preset
- Run
voice-dictation-setup if no clean preset is loaded yet
