WaveCap Audio Analysis Skill

Use this skill to analyze and inspect recorded audio files from WaveCap transcription sessions.

Recordings Location

Audio recordings are stored in the state directory:

STATE_DIR="/Users/thw/Projects/WaveCap/state"
RECORDINGS_DIR="$STATE_DIR/recordings"

Recordings are also accessible via HTTP at http://localhost:8000/recordings/

List Available Recordings

List all recordings

ls -la "$STATE_DIR/recordings/" | head -50

List recordings with sizes (sorted by date)

ls -lht "$STATE_DIR/recordings/" | head -30

Count total recordings

ls "$STATE_DIR/recordings/" 2>/dev/null | wc -l

Find recordings by date

# Today's recordings
find "$STATE_DIR/recordings" -name "*.wav" -mtime 0 -type f

# Last 7 days
find "$STATE_DIR/recordings" -name "*.wav" -mtime -7 -type f

Get Audio File Metadata

Basic file info

file "$STATE_DIR/recordings/FILENAME.wav"

Detailed audio properties (requires ffprobe/sox)

# Using ffprobe (if available)
ffprobe -v quiet -show_format -show_streams "$STATE_DIR/recordings/FILENAME.wav" 2>/dev/null | grep -E "duration|sample_rate|channels|bit_rate"

# Using soxi (if sox installed)
soxi "$STATE_DIR/recordings/FILENAME.wav" 2>/dev/null

Get duration of a recording

ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$STATE_DIR/recordings/FILENAME.wav" 2>/dev/null

Find Recordings for a Transcription

Get recording URL from transcription

curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?limit=10" | \
  jq '.transcriptions[] | {id, timestamp, recordingUrl, text}'

Check if recording file exists

RECORDING_URL="recordings/abc123.wav"
FILENAME=$(basename "$RECORDING_URL")
ls -la "$STATE_DIR/recordings/$FILENAME" 2>/dev/null || echo "Recording not found"

Analyze Audio Quality

Check for very short recordings (likely noise/errors)

find "$STATE_DIR/recordings" -name "*.wav" -size -10k -type f | head -20

Check for large recordings (long transmissions)

find "$STATE_DIR/recordings" -name "*.wav" -size +500k -type f | head -20

Get total disk usage

du -sh "$STATE_DIR/recordings/"

List recordings without transcriptions

# Get all transcription recording URLs
TRANSCRIBED=$(curl -s http://localhost:8000/api/transcriptions/export | \
  jq -r '.[].recordingUrl // empty' | xargs -I{} basename {} | sort | uniq)

# Compare with files on disk
ls "$STATE_DIR/recordings/" | sort > /tmp/all_recordings.txt
echo "$TRANSCRIBED" | sort > /tmp/transcribed.txt
comm -23 /tmp/all_recordings.txt /tmp/transcribed.txt | head -20

Stream Audio via API

Access recording via HTTP

# Recordings are served at /recordings/
curl -s -I "http://localhost:8000/recordings/FILENAME.wav"

# Download a recording
curl -o local_copy.wav "http://localhost:8000/recordings/FILENAME.wav"

Stream live audio (if stream supports it)

# Check if stream supports live audio
curl -s http://localhost:8000/api/streams | jq '.[] | select(.source == "audio" or .source == "remote") | {id, name, source}'

Clean Up Old Recordings

Find recordings older than 30 days

find "$STATE_DIR/recordings" -name "*.wav" -mtime +30 -type f | wc -l

Calculate space used by old recordings

find "$STATE_DIR/recordings" -name "*.wav" -mtime +30 -type f -exec du -ch {} + 2>/dev/null | tail -1

Tips

• Recording filenames typically include a UUID or timestamp
• Use recordingUrl from transcription API responses to link audio to transcripts
• The recordings directory may grow large over time - monitor disk usage
• Audio files are WAV format, typically 16kHz mono
• Very short recordings (< 1 second) may be noise or squelch breaks