Support alternative sample MIME types in MuxerWrapper

Enable transformer to transmux into alternative sample MIME types.
For example, some Dolby Vision profiles have a backwards-compatible
AVC or HEVC layer. MV-HEVC is backwards compatible with HEVC.

This change enables Transformer to transmux into the backwards compatible
format to improve compatibility with legacy APIs such as
MediaMetadataRetriever.

PiperOrigin-RevId: 693667597

RELEASENOTES.md

@@ -26,6 +26,8 @@
 *   Transformer:
     *   Update parameters of `VideoFrameProcessor.registerInputStream` and
         `VideoFrameProcessor.Listener.onInputStreamRegistered` to use `Format`.
+    *   Add support for transmuxing into alternative backwards compatible
+        formats.
 *   Extractors:
 *   DataSource:
     *   `DataSourceContractTest`: Assert that `DataSource.getUri()` and

libraries/test_data/src/test/assets/transformerdumps/mp4/dolbyVision-hdr.MOV/transmuxed_with_inappmuxer.dump

@@ -0,0 +1,96 @@
+seekMap:
+  isSeekable = true
+  duration = 166700
+  getPosition(0) = [[timeUs=0, position=400052]]
+  getPosition(1) = [[timeUs=0, position=400052]]
+  getPosition(83350) = [[timeUs=0, position=400052]]
+  getPosition(166700) = [[timeUs=0, position=400052]]
+numberOfTracks = 2
+track 0:
+  total output bytes = 56045
+  sample count = 5
+  track duration = 166700
+  format 0:
+    id = 1
+    sampleMimeType = video/hevc
+    codecs = hvc1.2.4.L93.B0
+    maxInputSize = 30755
+    maxNumReorderSamples = 2
+    width = 1280
+    height = 720
+    frameRate = 30.00
+    rotationDegrees = 90
+    colorInfo:
+      colorSpace = 6
+      colorRange = 2
+      colorTransfer = 7
+      lumaBitdepth = 10
+      chromaBitdepth = 10
+    metadata = entries=[mdta: key=com.apple.quicktime.location.accuracy.horizontal, value=35.000000, mdta: key=com.apple.quicktime.location.ISO6709, value=+51.5334-000.1255+026.756/, mdta: key=com.apple.quicktime.software, value=16.4.1, mdta: key=com.apple.quicktime.creationdate, value=2023-05-09T12:41:46+0100, mdta: key=com.apple.quicktime.make, value=Apple, mdta: key=com.apple.quicktime.model, value=iPhone 14 Pro, Mp4Timestamp: creation time=3000000000, modification time=4000000000, timescale=10000]
+    initializationData:
+      data = length 97, hash 71D67547
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 30725, hash A9F08085
+  sample 1:
+    time = 133333
+    flags = 0
+    data = length 11543, hash D0A79439
+  sample 2:
+    time = 66666
+    flags = 0
+    data = length 6139, hash 3FCEAFBE
+  sample 3:
+    time = 33333
+    flags = 0
+    data = length 3660, hash A80DE201
+  sample 4:
+    time = 100000
+    flags = 536870912
+    data = length 3978, hash 59521A2C
+track 1:
+  total output bytes = 2674
+  sample count = 7
+  track duration = 162600
+  format 0:
+    peakBitrate = 192000
+    id = 2
+    sampleMimeType = audio/mp4a-latm
+    codecs = mp4a.40.2
+    maxInputSize = 616
+    channelCount = 2
+    sampleRate = 44100
+    language = und
+    metadata = entries=[mdta: key=com.apple.quicktime.location.accuracy.horizontal, value=35.000000, mdta: key=com.apple.quicktime.location.ISO6709, value=+51.5334-000.1255+026.756/, mdta: key=com.apple.quicktime.software, value=16.4.1, mdta: key=com.apple.quicktime.creationdate, value=2023-05-09T12:41:46+0100, mdta: key=com.apple.quicktime.make, value=Apple, mdta: key=com.apple.quicktime.model, value=iPhone 14 Pro, Mp4Timestamp: creation time=3000000000, modification time=4000000000, timescale=10000]
+    initializationData:
+      data = length 2, hash 5FF
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 6, hash 6EFB71F1
+  sample 1:
+    time = 23229
+    flags = 1
+    data = length 260, hash EE23C4F4
+  sample 2:
+    time = 46458
+    flags = 1
+    data = length 458, hash C54D6937
+  sample 3:
+    time = 69687
+    flags = 1
+    data = length 448, hash 319412F3
+  sample 4:
+    time = 92916
+    flags = 1
+    data = length 422, hash 1E293713
+  sample 5:
+    time = 116145
+    flags = 1
+    data = length 586, hash C4CD7E3D
+  sample 6:
+    time = 139375
+    flags = 536870913
+    data = length 494, hash A7AFD1B8
+tracksEnded = true

libraries/transformer/src/androidTest/java/androidx/media3/transformer/TransformerEndToEndTest.java

@@ -25,6 +25,7 @@ import static androidx.media3.transformer.AndroidTestUtil.JPG_ASSET;
 import static androidx.media3.transformer.AndroidTestUtil.JPG_PIXEL_MOTION_PHOTO_ASSET;
 import static androidx.media3.transformer.AndroidTestUtil.MP3_ASSET;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET;
+import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_DOLBY_VISION_HDR;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_PHOTOS_TRIM_OPTIMIZATION_VIDEO;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_WITH_INCREASING_TIMESTAMPS;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_WITH_INCREASING_TIMESTAMPS_320W_240H_15S;
@@ -1773,6 +1774,60 @@ public class TransformerEndToEndTest {
     assertThat(new File(result.filePath).length()).isGreaterThan(0);
   }
 
+  @Test
+  public void dolbyVisionVideo_noEffects_transmuxesToHevc() throws Exception {
+    assumeTrue("This test requires B-frame support", Util.SDK_INT > 24);
+    assumeTrue(
+        new DefaultMuxer.Factory()
+            .getSupportedSampleMimeTypes(C.TRACK_TYPE_VIDEO)
+            .contains(MimeTypes.VIDEO_H265));
+    EditedMediaItem editedMediaItem =
+        new EditedMediaItem.Builder(MediaItem.fromUri(Uri.parse(MP4_ASSET_DOLBY_VISION_HDR.uri)))
+            .setRemoveAudio(true)
+            .build();
+
+    ExportTestResult result =
+        new TransformerAndroidTestRunner.Builder(
+                context,
+                new Transformer.Builder(context).setVideoMimeType(MimeTypes.VIDEO_H265).build())
+            .build()
+            .run(testId, editedMediaItem);
+
+    MediaExtractorCompat mediaExtractor = new MediaExtractorCompat(context);
+    mediaExtractor.setDataSource(Uri.parse(result.filePath), /* offset= */ 0);
+    checkState(mediaExtractor.getTrackCount() == 1);
+    MediaFormat mediaFormat = mediaExtractor.getTrackFormat(/* trackIndex= */ 0);
+    Format format = createFormatFromMediaFormat(mediaFormat);
+    assertThat(format.sampleMimeType).isEqualTo(MimeTypes.VIDEO_H265);
+    assertThat(result.exportResult.videoConversionProcess).isEqualTo(CONVERSION_PROCESS_TRANSMUXED);
+  }
+
+  @Test
+  public void dolbyVisionVideo_noEffects_withInAppMuxer_transmuxesToHevc() throws Exception {
+    EditedMediaItem editedMediaItem =
+        new EditedMediaItem.Builder(MediaItem.fromUri(Uri.parse(MP4_ASSET_DOLBY_VISION_HDR.uri)))
+            .setRemoveAudio(true)
+            .build();
+
+    ExportTestResult result =
+        new TransformerAndroidTestRunner.Builder(
+                context,
+                new Transformer.Builder(context)
+                    .setVideoMimeType(MimeTypes.VIDEO_H265)
+                    .setMuxerFactory(new InAppMuxer.Factory.Builder().build())
+                    .build())
+            .build()
+            .run(testId, editedMediaItem);
+
+    MediaExtractorCompat mediaExtractor = new MediaExtractorCompat(context);
+    mediaExtractor.setDataSource(Uri.parse(result.filePath), /* offset= */ 0);
+    checkState(mediaExtractor.getTrackCount() == 1);
+    MediaFormat mediaFormat = mediaExtractor.getTrackFormat(/* trackIndex= */ 0);
+    Format format = createFormatFromMediaFormat(mediaFormat);
+    assertThat(format.sampleMimeType).isEqualTo(MimeTypes.VIDEO_H265);
+    assertThat(result.exportResult.videoConversionProcess).isEqualTo(CONVERSION_PROCESS_TRANSMUXED);
+  }
+
   @Test
   public void audioComposition_compositionEffects_transcodes() throws Exception {
     EditedMediaItem editedMediaItem =

libraries/transformer/src/main/java/androidx/media3/transformer/SampleExporter.java

@@ -19,6 +19,7 @@ package androidx.media3.transformer;
 import static androidx.media3.common.ColorInfo.isTransferHdr;
 import static androidx.media3.common.util.Assertions.checkNotNull;
 import static androidx.media3.common.util.Assertions.checkStateNotNull;
+import static androidx.media3.exoplayer.mediacodec.MediaCodecUtil.getAlternativeCodecMimeType;
 import static androidx.media3.transformer.EncoderUtil.getSupportedEncoders;
 import static androidx.media3.transformer.EncoderUtil.getSupportedEncodersForHdrEditing;
 import static androidx.media3.transformer.TransformerUtil.getProcessedTrackType;
@@ -109,6 +110,13 @@ import java.util.List;
       if (metadata != null) {
         inputFormat = inputFormat.buildUpon().setMetadata(metadata).build();
       }
+      if (!muxerWrapper.supportsSampleMimeType(inputFormat.sampleMimeType)) {
+        String alternativeSampleMimeType = getAlternativeCodecMimeType(inputFormat);
+        if (muxerWrapper.supportsSampleMimeType(alternativeSampleMimeType)) {
+          inputFormat =
+              inputFormat.buildUpon().setSampleMimeType(alternativeSampleMimeType).build();
+        }
+      }
       try {
         muxerWrapper.addTrackFormat(inputFormat);
       } catch (MuxerException e) {

libraries/transformer/src/main/java/androidx/media3/transformer/TransformerUtil.java

@@ -19,6 +19,7 @@ package androidx.media3.transformer;
 import static androidx.media3.common.ColorInfo.SDR_BT709_LIMITED;
 import static androidx.media3.common.ColorInfo.isTransferHdr;
 import static androidx.media3.common.util.Assertions.checkArgument;
+import static androidx.media3.exoplayer.mediacodec.MediaCodecUtil.getAlternativeCodecMimeType;
 import static androidx.media3.transformer.Composition.HDR_MODE_KEEP_HDR;
 import static androidx.media3.transformer.Composition.HDR_MODE_TONE_MAP_HDR_TO_SDR_USING_OPEN_GL;
 import static androidx.media3.transformer.EncoderUtil.getSupportedEncodersForHdrEditing;
@@ -153,12 +154,18 @@ public final class TransformerUtil {
     if (transformationRequest.hdrMode != HDR_MODE_KEEP_HDR) {
       return true;
     }
-    if (transformationRequest.videoMimeType != null
+    @Nullable String requestedMimeType = transformationRequest.videoMimeType;
-        && !transformationRequest.videoMimeType.equals(inputFormat.sampleMimeType)) {
+    if (requestedMimeType != null) {
+      boolean requestedMimeTypeEqualsPrimaryOrAlternativeMimeType =
+          requestedMimeType.equals(inputFormat.sampleMimeType)
+              || requestedMimeType.equals(getAlternativeCodecMimeType(inputFormat));
+      if (!requestedMimeTypeEqualsPrimaryOrAlternativeMimeType) {
         return true;
       }
-    if (transformationRequest.videoMimeType == null
+    }
-        && !muxerWrapper.supportsSampleMimeType(inputFormat.sampleMimeType)) {
+    if (requestedMimeType == null
+        && !muxerWrapper.supportsSampleMimeType(inputFormat.sampleMimeType)
+        && !muxerWrapper.supportsSampleMimeType(getAlternativeCodecMimeType(inputFormat))) {
       return true;
     }
     if (inputFormat.pixelWidthHeightRatio != 1f) {

libraries/transformer/src/test/java/androidx/media3/transformer/TransformerWithInAppMuxerEndToEndParameterizedTest.java

@@ -53,6 +53,7 @@ public class TransformerWithInAppMuxerEndToEndParameterizedTest {
   private static final String AMR_WB_3GP = "mp4/bbb_mono_16kHz_23.05kbps_amrwb.3gp";
   private static final String OPUS_OGG = "mp4/bbb_6ch_8kHz_opus.ogg";
   private static final String VORBIS_OGG = "mp4/bbb_1ch_16kHz_q10_vorbis.ogg";
+  private static final String DOLBY_VISION_MOV = "mp4/dolbyVision-hdr.MOV";
 
   @Parameters(name = "{0}")
   public static ImmutableList<String> mediaFiles() {
@@ -65,7 +66,8 @@ public class TransformerWithInAppMuxerEndToEndParameterizedTest {
         AMR_NB_3GP,
         AMR_WB_3GP,
         OPUS_OGG,
-        VORBIS_OGG);
+        VORBIS_OGG,
+        DOLBY_VISION_MOV);
   }
 
   @Parameter public @MonotonicNonNull String inputFile;