tirtc-devtools-cli 0.3.0 → 0.3.2

package/vendor/runtime/script/prepare_runtime_media_dataset.sh

@@ -133,19 +133,19 @@ require_cmd "$FFPROBE_BIN"
 SOURCE_ABS="$(cd "$(dirname "$SOURCE")" && pwd)/$(basename "$SOURCE")"
 SOURCE_FILE_NAME="$(basename "$SOURCE_ABS")"
 
-readonly k_schema_version=2
-readonly k_tool_version="runtime_local_assets_prepare_v5"
+readonly k_schema_version=3
+readonly k_tool_version="runtime_local_assets_prepare_v6"
 readonly k_video_max_width=1280
 readonly k_video_max_height=720
 readonly k_video_fps=15
 readonly k_audio_sample_rate_hz=8000
 readonly k_audio_sample_rate_16k_hz=16000
-readonly k_audio_channels=1
 readonly k_audio_packet_duration_ms=40
+readonly k_audio_bits_per_sample=16
+readonly k_aac_packet_samples_per_channel=1024
 readonly k_mjpeg_packet_duration_ms=80
 readonly k_mjpeg_extract_interval_seconds=10
 readonly k_loop_policy="continuous_pts"
-readonly k_audio_codec="g711a"
 readonly k_output_audio_format="pcm_s16"
 readonly k_output_video_format="rgba8888"
 
@@ -253,10 +253,6 @@ VIDEO_H265_PATH="$ASSETS_DIR/video/video.h265"
 VIDEO_H265_PACKET_INDEX_PATH="$ASSETS_DIR/video/video_h265_packets.csv"
 VIDEO_MJPEG_PATH="$ASSETS_DIR/video/video.mjpeg"
 VIDEO_MJPEG_PACKET_INDEX_PATH="$ASSETS_DIR/video/video_mjpeg_packets.csv"
-AUDIO_PATH="$ASSETS_DIR/audio/audio.g711a"
-AUDIO_PACKET_INDEX_PATH="$ASSETS_DIR/audio/audio_packets.csv"
-AUDIO_16K_PATH="$ASSETS_DIR/audio/audio_16k.g711a"
-AUDIO_16K_PACKET_INDEX_PATH="$ASSETS_DIR/audio/audio_16k_packets.csv"
 
 video_encode_args=(
   -vf "fps=${k_video_fps},scale=${TARGET_VIDEO_WIDTH}:${TARGET_VIDEO_HEIGHT}:flags=lanczos"
@@ -335,26 +331,6 @@ if ! find "$mjpeg_frame_dir" -type f -name 'frame_*.jpg' | grep -q .; then
     "$mjpeg_frame_dir/frame_000001.jpg"
 fi
 
-progress "encoding 8k g711a audio track"
-"$FFMPEG_BIN" -hide_banner -loglevel error -y \
-  -i "$SOURCE_ABS" \
-  -vn \
-  -ac ${k_audio_channels} \
-  -ar ${k_audio_sample_rate_hz} \
-  -c:a pcm_alaw \
-  -f alaw \
-  "$AUDIO_PATH"
-
-progress "encoding 16k g711a audio track"
- "$FFMPEG_BIN" -hide_banner -loglevel error -y \
-  -i "$SOURCE_ABS" \
-  -vn \
-  -ac ${k_audio_channels} \
-  -ar ${k_audio_sample_rate_16k_hz} \
-  -c:a pcm_alaw \
-  -f alaw \
-  "$AUDIO_16K_PATH"
-
 progress "indexing h264 video packets"
 VIDEO_PACKET_INDEX_PATH="$VIDEO_H264_PACKET_INDEX_PATH" \
 VIDEO_PATH="$VIDEO_H264_PATH" \
@@ -607,16 +583,35 @@ for (let index = 0; index < packetCount; index += 1) {
 fs.writeFileSync(indexPath, `${lines.join('\n')}\n`, 'utf8');
 NODE
 
-progress "indexing 8k g711a audio packets"
-AUDIO_BYTES="$(wc -c < "$AUDIO_PATH" | tr -d '[:space:]')"
-AUDIO_PACKET_BYTES=$((k_audio_sample_rate_hz * k_audio_packet_duration_ms / 1000))
-[[ "$AUDIO_PACKET_BYTES" -gt 0 ]] || fail "invalid audio packet bytes"
-AUDIO_REMAINDER=$((AUDIO_BYTES % AUDIO_PACKET_BYTES))
-if [[ "$AUDIO_REMAINDER" -ne 0 ]]; then
-  AUDIO_ALIGNED_BYTES=$((AUDIO_BYTES - AUDIO_REMAINDER))
-  [[ "$AUDIO_ALIGNED_BYTES" -gt 0 ]] || fail "audio output is too short for one full packet"
-  echo "[prepare_runtime_media_dataset] trim trailing audio bytes to packet boundary: removed=${AUDIO_REMAINDER}" >&2
-  AUDIO_PATH="$AUDIO_PATH" AUDIO_ALIGNED_BYTES="$AUDIO_ALIGNED_BYTES" node <<'NODE'
+encode_g711a_track() {
+  local sample_rate="$1"
+  local channels="$2"
+  local key="$3"
+  local output_path="$ASSETS_DIR/audio/${key}.g711a"
+  local index_path="$ASSETS_DIR/audio/${key}.csv"
+
+  progress "encoding ${key} audio track"
+  "$FFMPEG_BIN" -hide_banner -loglevel error -y \
+    -i "$SOURCE_ABS" \
+    -vn \
+    -ac "$channels" \
+    -ar "$sample_rate" \
+    -c:a pcm_alaw \
+    -f alaw \
+    "$output_path"
+
+  progress "indexing ${key} audio packets"
+  local audio_bytes
+  audio_bytes="$(wc -c < "$output_path" | tr -d '[:space:]')"
+  local samples_per_packet=$((sample_rate * k_audio_packet_duration_ms / 1000))
+  local packet_bytes=$((samples_per_packet * channels))
+  [[ "$packet_bytes" -gt 0 ]] || fail "invalid audio packet bytes"
+  local remainder=$((audio_bytes % packet_bytes))
+  if [[ "$remainder" -ne 0 ]]; then
+    local aligned_bytes=$((audio_bytes - remainder))
+    [[ "$aligned_bytes" -gt 0 ]] || fail "${key} audio output is too short for one full packet"
+    echo "[prepare_runtime_media_dataset] trim trailing ${key} bytes to packet boundary: removed=${remainder}" >&2
+    AUDIO_PATH="$output_path" AUDIO_ALIGNED_BYTES="$aligned_bytes" node <<'NODE'
 const fs = require('fs');
 const audioPath = process.env.AUDIO_PATH;
 const alignedBytes = Number(process.env.AUDIO_ALIGNED_BYTES || '0');
@@ -625,51 +620,93 @@ if (!audioPath || !Number.isFinite(alignedBytes) || alignedBytes <= 0) {
 }
 fs.truncateSync(audioPath, alignedBytes);
 NODE
-  AUDIO_BYTES="$AUDIO_ALIGNED_BYTES"
-fi
-AUDIO_PACKET_COUNT=$((AUDIO_BYTES / AUDIO_PACKET_BYTES))
-[[ "$AUDIO_PACKET_COUNT" -gt 0 ]] || fail "generated audio packets are empty"
-
-{
-  echo "pts_us,offset,size"
-  for ((i = 0; i < AUDIO_PACKET_COUNT; i += 1)); do
-    pts_us=$((i * k_audio_packet_duration_ms * 1000))
-    offset=$((i * AUDIO_PACKET_BYTES))
-    echo "${pts_us},${offset},${AUDIO_PACKET_BYTES}"
-  done
-} > "$AUDIO_PACKET_INDEX_PATH"
-
-progress "indexing 16k g711a audio packets"
-AUDIO_16K_BYTES="$(wc -c < "$AUDIO_16K_PATH" | tr -d '[:space:]')"
-AUDIO_16K_PACKET_BYTES=$((k_audio_sample_rate_16k_hz * k_audio_packet_duration_ms / 1000))
-[[ "$AUDIO_16K_PACKET_BYTES" -gt 0 ]] || fail "invalid 16k audio packet bytes"
-AUDIO_16K_REMAINDER=$((AUDIO_16K_BYTES % AUDIO_16K_PACKET_BYTES))
-if [[ "$AUDIO_16K_REMAINDER" -ne 0 ]]; then
-  AUDIO_16K_ALIGNED_BYTES=$((AUDIO_16K_BYTES - AUDIO_16K_REMAINDER))
-  [[ "$AUDIO_16K_ALIGNED_BYTES" -gt 0 ]] || fail "16k audio output is too short for one full packet"
-  echo "[prepare_runtime_media_dataset] trim trailing 16k audio bytes to packet boundary: removed=${AUDIO_16K_REMAINDER}" >&2
-  AUDIO_PATH="$AUDIO_16K_PATH" AUDIO_ALIGNED_BYTES="$AUDIO_16K_ALIGNED_BYTES" node <<'NODE'
+    audio_bytes="$aligned_bytes"
+  fi
+  local packet_count=$((audio_bytes / packet_bytes))
+  [[ "$packet_count" -gt 0 ]] || fail "generated ${key} audio packets are empty"
+
+  {
+    echo "pts_us,offset,size,samples_per_channel"
+    for ((i = 0; i < packet_count; i += 1)); do
+      pts_us=$((i * k_audio_packet_duration_ms * 1000))
+      offset=$((i * packet_bytes))
+      echo "${pts_us},${offset},${packet_bytes},${samples_per_packet}"
+    done
+  } > "$index_path"
+}
+
+encode_aac_track() {
+  local sample_rate="$1"
+  local channels="$2"
+  local key="$3"
+  local adts_path="$ASSETS_DIR/audio/${key}.adts"
+  local output_path="$ASSETS_DIR/audio/${key}.aac"
+  local index_path="$ASSETS_DIR/audio/${key}.csv"
+
+  progress "encoding ${key} audio track"
+  "$FFMPEG_BIN" -hide_banner -loglevel error -y \
+    -i "$SOURCE_ABS" \
+    -vn \
+    -ac "$channels" \
+    -ar "$sample_rate" \
+    -c:a aac \
+    -profile:a aac_low \
+    -f adts \
+    "$adts_path"
+
+  progress "indexing ${key} audio packets"
+  ADTS_PATH="$adts_path" OUTPUT_PATH="$output_path" INDEX_PATH="$index_path" SAMPLE_RATE="$sample_rate" AAC_PACKET_SAMPLES_PER_CHANNEL="$k_aac_packet_samples_per_channel" node <<'NODE'
 const fs = require('fs');
-const audioPath = process.env.AUDIO_PATH;
-const alignedBytes = Number(process.env.AUDIO_ALIGNED_BYTES || '0');
-if (!audioPath || !Number.isFinite(alignedBytes) || alignedBytes <= 0) {
-  throw new Error('audio trim env is incomplete');
+const adtsPath = process.env.ADTS_PATH;
+const outputPath = process.env.OUTPUT_PATH;
+const indexPath = process.env.INDEX_PATH;
+const sampleRate = Number(process.env.SAMPLE_RATE || '0');
+if (!adtsPath || !outputPath || !indexPath || !Number.isFinite(sampleRate) || sampleRate <= 0) {
+  throw new Error('aac packet index env is incomplete');
+}
+const data = fs.readFileSync(adtsPath);
+const chunks = [];
+const rows = ['pts_us,offset,size,samples_per_channel'];
+let cursor = 0;
+let outputOffset = 0;
+let packetIndex = 0;
+const samplesPerChannel = Number(process.env.AAC_PACKET_SAMPLES_PER_CHANNEL || '1024');
+if (!Number.isFinite(samplesPerChannel) || samplesPerChannel <= 0) {
+  throw new Error('aac packet samples env is invalid');
+}
+while (cursor + 7 <= data.length) {
+  if (data[cursor] !== 0xff || (data[cursor + 1] & 0xf0) !== 0xf0) {
+    throw new Error(`aac adts sync missing at offset ${cursor}`);
+  }
+  const protectionAbsent = data[cursor + 1] & 0x01;
+  const headerBytes = protectionAbsent ? 7 : 9;
+  const frameLength = ((data[cursor + 3] & 0x03) << 11) | (data[cursor + 4] << 3) | ((data[cursor + 5] & 0xe0) >> 5);
+  if (frameLength <= headerBytes || cursor + frameLength > data.length) {
+    throw new Error(`aac adts invalid frame length at offset ${cursor}`);
+  }
+  const payload = data.subarray(cursor + headerBytes, cursor + frameLength);
+  const ptsUs = Math.round(packetIndex * samplesPerChannel * 1_000_000 / sampleRate);
+  rows.push(`${ptsUs},${outputOffset},${payload.length},${samplesPerChannel}`);
+  chunks.push(payload);
+  outputOffset += payload.length;
+  packetIndex += 1;
+  cursor += frameLength;
 }
-fs.truncateSync(audioPath, alignedBytes);
+if (packetIndex === 0 || outputOffset === 0) {
+  throw new Error('aac packet index is empty');
+}
+fs.writeFileSync(outputPath, Buffer.concat(chunks));
+fs.writeFileSync(indexPath, `${rows.join('\n')}\n`, 'utf8');
+fs.rmSync(adtsPath, {force: true});
 NODE
-  AUDIO_16K_BYTES="$AUDIO_16K_ALIGNED_BYTES"
-fi
-AUDIO_16K_PACKET_COUNT=$((AUDIO_16K_BYTES / AUDIO_16K_PACKET_BYTES))
-[[ "$AUDIO_16K_PACKET_COUNT" -gt 0 ]] || fail "generated 16k audio packets are empty"
+}
 
-{
-  echo "pts_us,offset,size"
-  for ((i = 0; i < AUDIO_16K_PACKET_COUNT; i += 1)); do
-    pts_us=$((i * k_audio_packet_duration_ms * 1000))
-    offset=$((i * AUDIO_16K_PACKET_BYTES))
-    echo "${pts_us},${offset},${AUDIO_16K_PACKET_BYTES}"
+for sample_rate in "$k_audio_sample_rate_hz" "$k_audio_sample_rate_16k_hz"; do
+  for channels in 1 2; do
+    encode_g711a_track "$sample_rate" "$channels" "g711a_${sample_rate}_${channels}ch_s16"
+    encode_aac_track "$sample_rate" "$channels" "aac_${sample_rate}_${channels}ch_s16"
   done
-} > "$AUDIO_16K_PACKET_INDEX_PATH"
+done
 
 progress "writing prepared media manifest"
 cat > "$MANIFEST_PATH" <<JSON
@@ -690,12 +727,14 @@ cat > "$MANIFEST_PATH" <<JSON
     "bitrate_kbps": ${TARGET_VIDEO_BITRATE_KBPS}
   },
   "audio": {
-    "path": "audio/audio.g711a",
-    "packet_index_path": "audio/audio_packets.csv",
-    "codec": "${k_audio_codec}",
+    "path": "audio/g711a_8000_1ch_s16.g711a",
+    "packet_index_path": "audio/g711a_8000_1ch_s16.csv",
+    "codec": "g711a",
     "sample_rate_hz": ${k_audio_sample_rate_hz},
-    "channels": ${k_audio_channels},
-    "packet_duration_ms": ${k_audio_packet_duration_ms}
+    "channels": 1,
+    "bits_per_sample": ${k_audio_bits_per_sample},
+    "packet_duration_ms": ${k_audio_packet_duration_ms},
+    "packet_samples_per_channel": $((k_audio_sample_rate_hz * k_audio_packet_duration_ms / 1000))
   },
   "output": {
     "audio_format": "${k_output_audio_format}",
@@ -731,21 +770,85 @@ cat > "$MANIFEST_PATH" <<JSON
     }
   },
   "audio_tracks": {
-    "g711a_8k": {
-      "path": "audio/audio.g711a",
-      "packet_index_path": "audio/audio_packets.csv",
-      "codec": "${k_audio_codec}",
+    "g711a_8000_1ch_s16": {
+      "path": "audio/g711a_8000_1ch_s16.g711a",
+      "packet_index_path": "audio/g711a_8000_1ch_s16.csv",
+      "codec": "g711a",
+      "sample_rate_hz": ${k_audio_sample_rate_hz},
+      "channels": 1,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": ${k_audio_packet_duration_ms},
+      "packet_samples_per_channel": $((k_audio_sample_rate_hz * k_audio_packet_duration_ms / 1000))
+    },
+    "g711a_8000_2ch_s16": {
+      "path": "audio/g711a_8000_2ch_s16.g711a",
+      "packet_index_path": "audio/g711a_8000_2ch_s16.csv",
+      "codec": "g711a",
       "sample_rate_hz": ${k_audio_sample_rate_hz},
-      "channels": ${k_audio_channels},
-      "packet_duration_ms": ${k_audio_packet_duration_ms}
+      "channels": 2,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": ${k_audio_packet_duration_ms},
+      "packet_samples_per_channel": $((k_audio_sample_rate_hz * k_audio_packet_duration_ms / 1000))
+    },
+    "g711a_16000_1ch_s16": {
+      "path": "audio/g711a_16000_1ch_s16.g711a",
+      "packet_index_path": "audio/g711a_16000_1ch_s16.csv",
+      "codec": "g711a",
+      "sample_rate_hz": ${k_audio_sample_rate_16k_hz},
+      "channels": 1,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": ${k_audio_packet_duration_ms},
+      "packet_samples_per_channel": $((k_audio_sample_rate_16k_hz * k_audio_packet_duration_ms / 1000))
+    },
+    "g711a_16000_2ch_s16": {
+      "path": "audio/g711a_16000_2ch_s16.g711a",
+      "packet_index_path": "audio/g711a_16000_2ch_s16.csv",
+      "codec": "g711a",
+      "sample_rate_hz": ${k_audio_sample_rate_16k_hz},
+      "channels": 2,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": ${k_audio_packet_duration_ms},
+      "packet_samples_per_channel": $((k_audio_sample_rate_16k_hz * k_audio_packet_duration_ms / 1000))
+    },
+    "aac_8000_1ch_s16": {
+      "path": "audio/aac_8000_1ch_s16.aac",
+      "packet_index_path": "audio/aac_8000_1ch_s16.csv",
+      "codec": "aac",
+      "sample_rate_hz": ${k_audio_sample_rate_hz},
+      "channels": 1,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": $((k_aac_packet_samples_per_channel * 1000 / k_audio_sample_rate_hz)),
+      "packet_samples_per_channel": ${k_aac_packet_samples_per_channel}
+    },
+    "aac_8000_2ch_s16": {
+      "path": "audio/aac_8000_2ch_s16.aac",
+      "packet_index_path": "audio/aac_8000_2ch_s16.csv",
+      "codec": "aac",
+      "sample_rate_hz": ${k_audio_sample_rate_hz},
+      "channels": 2,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": $((k_aac_packet_samples_per_channel * 1000 / k_audio_sample_rate_hz)),
+      "packet_samples_per_channel": ${k_aac_packet_samples_per_channel}
+    },
+    "aac_16000_1ch_s16": {
+      "path": "audio/aac_16000_1ch_s16.aac",
+      "packet_index_path": "audio/aac_16000_1ch_s16.csv",
+      "codec": "aac",
+      "sample_rate_hz": ${k_audio_sample_rate_16k_hz},
+      "channels": 1,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": $((k_aac_packet_samples_per_channel * 1000 / k_audio_sample_rate_16k_hz)),
+      "packet_samples_per_channel": ${k_aac_packet_samples_per_channel}
     },
-    "g711a_16k": {
-      "path": "audio/audio_16k.g711a",
-      "packet_index_path": "audio/audio_16k_packets.csv",
-      "codec": "${k_audio_codec}",
+    "aac_16000_2ch_s16": {
+      "path": "audio/aac_16000_2ch_s16.aac",
+      "packet_index_path": "audio/aac_16000_2ch_s16.csv",
+      "codec": "aac",
       "sample_rate_hz": ${k_audio_sample_rate_16k_hz},
-      "channels": ${k_audio_channels},
-      "packet_duration_ms": ${k_audio_packet_duration_ms}
+      "channels": 2,
+      "bits_per_sample": ${k_audio_bits_per_sample},
+      "packet_duration_ms": $((k_aac_packet_samples_per_channel * 1000 / k_audio_sample_rate_16k_hz)),
+      "packet_samples_per_channel": ${k_aac_packet_samples_per_channel}
     }
   },
   "loop_policy": "${k_loop_policy}"