Support MKV embedded WebVTT captions

6b3187cc · Marcel Dopita · b3981be8 · 6b3187cc
Commit 6b3187cc authored Jan 15, 2022 by Marcel Dopita
Showing with 45 additions and 2 deletions
library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mkv/MatroskaExtractor.java
--- a/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mkv/MatroskaExtractor.java
+++ b/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mkv/MatroskaExtractor.java
@@ -138,6 +138,7 @@ public class MatroskaExtractor implements Extractor {
  private static final String CODEC_ID_PCM_FLOAT = "A_PCM/FLOAT/IEEE";
  private static final String CODEC_ID_SUBRIP = "S_TEXT/UTF8";
  private static final String CODEC_ID_ASS = "S_TEXT/ASS";
+  private static final String CODEC_ID_VTT = "S_TEXT/WEBVTT";
  private static final String CODEC_ID_VOBSUB = "S_VOBSUB";
  private static final String CODEC_ID_PGS = "S_HDMV/PGS";
  private static final String CODEC_ID_DVBSUB = "S_DVBSUB";
@@ -323,6 +324,32 @@ public class MatroskaExtractor implements Extractor {
  /** The format of an SSA timecode. */
  private static final String SSA_TIMECODE_FORMAT = "%01d:%02d:%02d:%02d";
+  /**
+   * A template for the prefix that must be added to each VTT sample.
+   *
+   * <p>The display time of each subtitle is passed as {@code timeUs} to {@link
+   * TrackOutput#sampleMetadata}. The start and end timecodes in this template are relative to
+   * {@code timeUs}. Hence the start timecode is always zero. The 12 byte end timecode starting at
+   * {@link #VTT_PREFIX_END_TIMECODE_OFFSET} is set to a placeholder value, and must be replaced
+   * with the duration of the subtitle.
+   *
+   * <p>Equivalent to the UTF-8 string: "WEBVTT\n\n00:00:00.000 --> 00:00:00.000\n".
+   */
+  private static final byte[] VTT_PREFIX =
+      new byte[]{
+          87, 69, 66, 86, 84, 84, 10, 10, 48, 48, 58, 48, 48, 58, 48, 48, 46, 48, 48, 48, 32, 45,
+          45, 62, 32, 48, 48, 58, 48, 48, 58, 48, 48, 46, 48, 48, 48, 10
+      };
+  /** The byte offset of the end timecode in {@link #VTT_PREFIX}. */
+  private static final int VTT_PREFIX_END_TIMECODE_OFFSET = 25;
+  /**
+   * The value by which to divide a time in microseconds to convert it to the unit of the last value
+   * in a VTT timecode (milliseconds).
+   */
+  private static final long VTT_TIMECODE_LAST_VALUE_SCALING_FACTOR = 1000;
+  /** The format of a VTT timecode. */
+  private static final String VTT_TIMECODE_FORMAT = "%02d:%02d:%02d.%03d";
  /** The length in bytes of a WAVEFORMATEX structure. */
  private static final int WAVE_FORMAT_SIZE = 18;
  /** Format tag indicating a WAVEFORMATEXTENSIBLE structure. */
@@ -1342,7 +1369,8 @@ public class MatroskaExtractor implements Extractor {
      track.trueHdSampleRechunker.sampleMetadata(
          track.output, timeUs, flags, size, offset, track.cryptoData);
    } else {
-      if (CODEC_ID_SUBRIP.equals(track.codecId) || CODEC_ID_ASS.equals(track.codecId)) {
+      if (CODEC_ID_SUBRIP.equals(track.codecId) || CODEC_ID_ASS.equals(track.codecId)
+          || CODEC_ID_VTT.equals(track.codecId)) {
        if (blockSampleCount > 1) {
          Log.w(TAG, "Skipping subtitle sample in laced block.");
        } else if (blockDurationUs == C.TIME_UNSET) {
@@ -1415,6 +1443,9 @@ public class MatroskaExtractor implements Extractor {
    } else if (CODEC_ID_ASS.equals(track.codecId)) {
      writeSubtitleSampleData(input, SSA_PREFIX, size);
      return finishWriteSampleData();
+    } else if (CODEC_ID_VTT.equals(track.codecId)) {
+      writeSubtitleSampleData(input, VTT_PREFIX, size);
+      return finishWriteSampleData();
    }
    TrackOutput output = track.output;
@@ -1641,7 +1672,8 @@ public class MatroskaExtractor implements Extractor {
   * <p>See documentation on {@link #SSA_DIALOGUE_FORMAT} and {@link #SUBRIP_PREFIX} for why we use
   * the duration as the end timecode.
   *
-   * @param codecId The subtitle codec; must be {@link #CODEC_ID_SUBRIP} or {@link #CODEC_ID_ASS}.
+   * @param codecId The subtitle codec; must be {@link #CODEC_ID_SUBRIP}, {@link #CODEC_ID_ASS} or
+   * {@link #CODEC_ID_VTT}.
   * @param durationUs The duration of the sample, in microseconds.
   * @param subtitleData The subtitle sample in which to overwrite the end timecode (output
   *     parameter).
@@ -1662,6 +1694,12 @@ public class MatroskaExtractor implements Extractor {
                durationUs, SSA_TIMECODE_FORMAT, SSA_TIMECODE_LAST_VALUE_SCALING_FACTOR);
        endTimecodeOffset = SSA_PREFIX_END_TIMECODE_OFFSET;
        break;
+      case CODEC_ID_VTT:
+        endTimecode =
+            formatSubtitleTimecode(
+                durationUs, VTT_TIMECODE_FORMAT, VTT_TIMECODE_LAST_VALUE_SCALING_FACTOR);
+        endTimecodeOffset = VTT_PREFIX_END_TIMECODE_OFFSET;
+        break;
      default:
        throw new IllegalArgumentException();
    }
@@ -1830,6 +1868,7 @@ public class MatroskaExtractor implements Extractor {
      case CODEC_ID_PCM_FLOAT:
      case CODEC_ID_SUBRIP:
      case CODEC_ID_ASS:
+      case CODEC_ID_VTT:
      case CODEC_ID_VOBSUB:
      case CODEC_ID_PGS:
      case CODEC_ID_DVBSUB:
@@ -2157,6 +2196,9 @@ public class MatroskaExtractor implements Extractor {
          mimeType = MimeTypes.TEXT_SSA;
          initializationData = ImmutableList.of(SSA_DIALOGUE_FORMAT, getCodecPrivate(codecId));
          break;
+        case CODEC_ID_VTT:
+          mimeType = MimeTypes.TEXT_VTT;
+          break;
        case CODEC_ID_VOBSUB:
          mimeType = MimeTypes.APPLICATION_VOBSUB;
          initializationData = ImmutableList.of(getCodecPrivate(codecId));
@@ -2245,6 +2287,7 @@ public class MatroskaExtractor implements Extractor {
            .setColorInfo(colorInfo);
      } else if (MimeTypes.APPLICATION_SUBRIP.equals(mimeType)
          || MimeTypes.TEXT_SSA.equals(mimeType)
+          || MimeTypes.TEXT_VTT.equals(mimeType)
          || MimeTypes.APPLICATION_VOBSUB.equals(mimeType)
          || MimeTypes.APPLICATION_PGS.equals(mimeType)
          || MimeTypes.APPLICATION_DVBSUBS.equals(mimeType)) {