1663b008ed
- Implemented helper functions for configuring Smart Mosaic layout, including setting participant count and waiting for participant visibility. - Created a new file for managing fake participants, allowing for joining and disconnecting from LiveKit rooms using both CLI and browser-based methods. - Introduced interfaces for browser-based fake participant options to streamline participant creation with audio and video assets.
331 lines
16 KiB
Bash
Executable File
331 lines
16 KiB
Bash
Executable File
#!/bin/bash
|
||
|
||
# =============================================================================
|
||
# Audio Generation Script for Smart Mosaic Layout Tests
|
||
# =============================================================================
|
||
# This script generates test audio files from a base audio file (base.wav)
|
||
# for testing the Smart Mosaic layout speaker detection functionality.
|
||
#
|
||
# Requirements:
|
||
# - ffmpeg 7.0+ (optimized for this version)
|
||
# - base.wav file with continuous speech audio in the same directory
|
||
#
|
||
# IMPORTANT: This script generates WAV files for best compatibility with
|
||
# Chrome's fake audio capture (--use-file-for-fake-audio-capture).
|
||
# WAV format ensures proper audio device simulation and VAD detection.
|
||
#
|
||
# Usage:
|
||
# chmod +x generate-test-audio.sh
|
||
# ./generate-test-audio.sh
|
||
# =============================================================================
|
||
|
||
set -e
|
||
|
||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||
BASE_AUDIO="$SCRIPT_DIR/base.wav"
|
||
OUTPUT_DIR="$SCRIPT_DIR"
|
||
|
||
# Audio settings
|
||
SAMPLE_RATE=48000
|
||
CHANNELS=1
|
||
|
||
# WAV encoding settings for Chrome fake audio capture compatibility
|
||
# PCM 16-bit is the most compatible format for Chrome's fake devices
|
||
WAV_OPTS="-c:a pcm_s16le -ar ${SAMPLE_RATE} -ac ${CHANNELS}"
|
||
|
||
# Check ffmpeg version
|
||
FFMPEG_VERSION=$(ffmpeg -version | head -n1 | grep -oP 'ffmpeg version \K[0-9]+')
|
||
echo "🔧 Detected ffmpeg major version: $FFMPEG_VERSION"
|
||
|
||
# Check if base audio exists
|
||
if [ ! -f "$BASE_AUDIO" ]; then
|
||
echo "❌ Error: base.wav not found in $SCRIPT_DIR"
|
||
echo "Please provide a base.wav file with continuous speech audio."
|
||
exit 1
|
||
fi
|
||
|
||
echo ""
|
||
echo "🎵 Generating test audio files from base.wav..."
|
||
echo " Output directory: $OUTPUT_DIR"
|
||
echo " Sample rate: ${SAMPLE_RATE}Hz, Channels: ${CHANNELS}"
|
||
echo " Codec: PCM 16-bit (WAV) for Chrome fake audio compatibility"
|
||
echo ""
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 1. continuous_speech.wav (30s)
|
||
# Continuous speech audio for participants who speak constantly
|
||
# -----------------------------------------------------------------------------
|
||
echo "1️⃣ Generating continuous_speech.wav (30s of continuous speech)..."
|
||
ffmpeg -y -i "$BASE_AUDIO" -t 30 -af "aresample=${SAMPLE_RATE}" $WAV_OPTS "$OUTPUT_DIR/continuous_speech.wav" 2>/dev/null
|
||
echo " ✅ continuous_speech.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 2. complete_silence.wav (30s)
|
||
# Complete digital silence using aevalsrc with explicit zero expression
|
||
# This generates samples with value exactly 0.0 - guaranteed no VAD trigger
|
||
# -----------------------------------------------------------------------------
|
||
echo "2️⃣ Generating complete_silence.wav (30s of TRUE digital silence)..."
|
||
ffmpeg -y -f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
$WAV_OPTS "$OUTPUT_DIR/complete_silence.wav" 2>/dev/null
|
||
echo " ✅ complete_silence.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 3. speech_5s_then_silence.wav (30s)
|
||
# 5s speech, then 25s TRUE silence
|
||
# Uses amix to combine speech with silence background for clean transitions
|
||
# -----------------------------------------------------------------------------
|
||
echo "3️⃣ Generating speech_5s_then_silence.wav (5s speech + 25s TRUE silence)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:5,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE}[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speech_5s_then_silence.wav" 2>/dev/null
|
||
echo " ✅ speech_5s_then_silence.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 4. silence_5s_then_speech.wav (30s)
|
||
# 5s TRUE silence, then 25s speech
|
||
# -----------------------------------------------------------------------------
|
||
echo "4️⃣ Generating silence_5s_then_speech.wav (5s TRUE silence + 25s speech)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:25,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=5s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/silence_5s_then_speech.wav" 2>/dev/null
|
||
echo " ✅ silence_5s_then_speech.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 5. speech_gap_speech.wav (30s)
|
||
# 5s speech, 10s TRUE silence, 15s speech - for testing speaker re-activation
|
||
# -----------------------------------------------------------------------------
|
||
echo "5️⃣ Generating speech_gap_speech.wav (5s speech + 10s TRUE gap + 15s speech)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:5,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE}[s1];
|
||
[0:a]atrim=5:20,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=15s:all=1[s2];
|
||
[1:a][s1][s2]amix=inputs=3:duration=first:dropout_transition=0,volume=3[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speech_gap_speech.wav" 2>/dev/null
|
||
echo " ✅ speech_gap_speech.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 6-11. Sequential speaker audio files (for rotation tests)
|
||
# Each speaker has a unique time window for speech with TRUE silence elsewhere
|
||
# -----------------------------------------------------------------------------
|
||
echo "6️⃣ Generating sequential speaker audio files (A through F)..."
|
||
|
||
# Speaker A: speaks 0-3s, then TRUE silence
|
||
echo " → speaker_seq_A.wav (speaks at 0-3s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE}[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_A.wav" 2>/dev/null
|
||
|
||
# Speaker B: TRUE silence 0-5s, speaks 5-8s, then TRUE silence
|
||
echo " → speaker_seq_B.wav (speaks at 5-8s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=5s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_B.wav" 2>/dev/null
|
||
|
||
# Speaker C: TRUE silence 0-10s, speaks 10-13s, then TRUE silence
|
||
echo " → speaker_seq_C.wav (speaks at 10-13s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=10s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_C.wav" 2>/dev/null
|
||
|
||
# Speaker D: TRUE silence 0-15s, speaks 15-18s, then TRUE silence
|
||
echo " → speaker_seq_D.wav (speaks at 15-18s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=15s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_D.wav" 2>/dev/null
|
||
|
||
# Speaker E: TRUE silence 0-20s, speaks 20-23s, then TRUE silence
|
||
echo " → speaker_seq_E.wav (speaks at 20-23s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=20s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_E.wav" 2>/dev/null
|
||
|
||
# Speaker F: TRUE silence 0-25s, speaks 25-28s, then TRUE silence
|
||
echo " → speaker_seq_F.wav (speaks at 25-28s)"
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:3,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=25s:all=1[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/speaker_seq_F.wav" 2>/dev/null
|
||
|
||
echo " ✅ Sequential speaker files created (A-F)"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 12. simultaneous_then_solo.wav (30s)
|
||
# 15s speech then 15s TRUE silence
|
||
# Used for the "simultaneous speech" test (this participant continues speaking)
|
||
# -----------------------------------------------------------------------------
|
||
echo "7️⃣ Generating simultaneous_then_solo.wav (15s speech + 15s TRUE silence)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:15,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE}[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/simultaneous_then_solo.wav" 2>/dev/null
|
||
echo " ✅ simultaneous_then_solo.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 13. simultaneous_then_stop.wav (30s)
|
||
# 5s speech then 25s TRUE silence
|
||
# Used for participants who stop speaking after simultaneous period
|
||
# -----------------------------------------------------------------------------
|
||
echo "8️⃣ Generating simultaneous_then_stop.wav (5s speech + 25s TRUE silence)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:5,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE}[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/simultaneous_then_stop.wav" 2>/dev/null
|
||
echo " ✅ simultaneous_then_stop.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 14. low_volume_speech.wav (30s)
|
||
# Continuous speech at 10% volume - below the audioLevel threshold (0.15)
|
||
# Used to test that participants with low audio levels are filtered out
|
||
# -----------------------------------------------------------------------------
|
||
echo "9️⃣ Generating low_volume_speech.wav (30s speech at 10% volume)..."
|
||
|
||
ffmpeg -y \
|
||
-f lavfi -i "anoisesrc=color=pink:amplitude=0.02:s=${SAMPLE_RATE}:d=30" \
|
||
$WAV_OPTS "$OUTPUT_DIR/ambient_pink_noise.wav" 2>/dev/null
|
||
|
||
echo " ✅ low_volume_speech.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 15. brief_sound_1s.wav (30s)
|
||
# Only 1 second of speech followed by silence
|
||
# Used to test minimum speaking duration filter (should be filtered out)
|
||
# -----------------------------------------------------------------------------
|
||
echo "🔟 Generating brief_sound_1s.wav (1s speech + 29s silence)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:1,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=5000|5000[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/brief_sound_1s_at_5s.wav" 2>/dev/null
|
||
echo " ✅ brief_sound_1s_at_5s.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# 16. brief_cough.wav (30s)
|
||
# Only 0.5 seconds of sound (simulating a cough) followed by silence
|
||
# Used to test that very brief sounds are filtered out
|
||
# -----------------------------------------------------------------------------
|
||
echo "1️⃣1️⃣ Generating brief_cough.wav (0.5s sound + 29.5s silence)..."
|
||
ffmpeg -y \
|
||
-i "$BASE_AUDIO" \
|
||
-f lavfi -i "aevalsrc=0:c=mono:s=${SAMPLE_RATE}:d=30" \
|
||
-filter_complex "
|
||
[0:a]atrim=0:0.5,asetpts=PTS-STARTPTS,aresample=${SAMPLE_RATE},adelay=5000|5000[speech];
|
||
[1:a][speech]amix=inputs=2:duration=first:dropout_transition=0,volume=2[out]
|
||
" \
|
||
-map "[out]" -t 30 $WAV_OPTS "$OUTPUT_DIR/brief_cough_at_5s.wav" 2>/dev/null
|
||
echo " ✅ brief_cough_at_5s.wav created"
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# Verify silence in generated files
|
||
# -----------------------------------------------------------------------------
|
||
echo ""
|
||
echo "🔍 Verifying silence quality in generated files..."
|
||
verify_silence() {
|
||
local file=$1
|
||
local expected_silence_start=$2
|
||
|
||
# Check RMS level in silence portion (should be exactly 0 or very close)
|
||
local rms=$(ffmpeg -i "$file" -af "atrim=${expected_silence_start}:${expected_silence_start}+1,astats=metadata=1:reset=1" -f null - 2>&1 | grep "RMS level" | head -1 | grep -oP '[-0-9.]+' | head -1)
|
||
|
||
if [ -n "$rms" ]; then
|
||
echo " $file: RMS at ${expected_silence_start}s = ${rms}dB"
|
||
fi
|
||
}
|
||
|
||
# Verify a few key files
|
||
verify_silence "$OUTPUT_DIR/complete_silence.wav" 15
|
||
verify_silence "$OUTPUT_DIR/speech_5s_then_silence.wav" 10
|
||
verify_silence "$OUTPUT_DIR/speaker_seq_B.wav" 2
|
||
|
||
# -----------------------------------------------------------------------------
|
||
# Summary
|
||
# -----------------------------------------------------------------------------
|
||
echo ""
|
||
echo "============================================================================="
|
||
echo "✅ Audio generation complete! (WAV format for Chrome fake audio capture)"
|
||
echo "============================================================================="
|
||
echo ""
|
||
echo "Generated files:"
|
||
echo " 📁 $OUTPUT_DIR/"
|
||
echo " ├── continuous_speech.wav (30s continuous speech)"
|
||
echo " ├── complete_silence.wav (30s TRUE digital silence - aevalsrc=0)"
|
||
echo " ├── speech_5s_then_silence.wav (5s speech + 25s TRUE silence)"
|
||
echo " ├── silence_5s_then_speech.wav (5s TRUE silence + 25s speech)"
|
||
echo " ├── speech_gap_speech.wav (5s speech + 10s gap + 15s speech)"
|
||
echo " ├── speaker_seq_A.wav (speaks at 0-3s)"
|
||
echo " ├── speaker_seq_B.wav (speaks at 5-8s)"
|
||
echo " ├── speaker_seq_C.wav (speaks at 10-13s)"
|
||
echo " ├── speaker_seq_D.wav (speaks at 15-18s)"
|
||
echo " ├── speaker_seq_E.wav (speaks at 20-23s)"
|
||
echo " ├── speaker_seq_F.wav (speaks at 25-28s)"
|
||
echo " ├── simultaneous_then_solo.wav (15s speech + 15s silence)"
|
||
echo " ├── simultaneous_then_stop.wav (5s speech + 25s silence)"
|
||
echo " ├── low_volume_speech.wav (30s speech at 10% volume - below threshold)"
|
||
echo " ├── brief_sound_1s.wav (1s speech + 29s silence - too short)"
|
||
echo " └── brief_cough.wav (0.5s sound + 29.5s silence - simulates cough)"
|
||
echo ""
|
||
echo "Key features of this version:"
|
||
echo " • WAV format (PCM 16-bit) for Chrome fake audio capture compatibility"
|
||
echo " • Uses aevalsrc=0 for TRUE digital silence (samples = 0.0)"
|
||
echo " • amix filter for clean speech/silence transitions"
|
||
echo " • adelay for precise speech timing"
|
||
echo " • 48kHz sample rate, mono channel"
|
||
echo ""
|
||
echo "Usage in tests:"
|
||
echo " await joinBrowserFakeParticipant(browser, roomId, 'speaker1', {"
|
||
echo " audioFile: 'continuous_speech.wav'"
|
||
echo " });"
|
||
echo ""
|