Throw exception when B-frames are written to Mp4Muxer

Mp4Muxer does not support out of order B-frames. Currently it
silently writes out of order B-frames, producing an invalid file (with
negative sample durations).

Although `Mp4Extractor` is somehow able to process this invalid file and
`Exoplayer` is able to play it but that is unexpected.

The `sample.mp4` test file contains B frames. Other test files does not
contain `H264 video + AAC audio` format hence created a new test file by
running `sample.mp4` via `Transformer` after applying some effects.

PiperOrigin-RevId: 594016144

libraries/muxer/src/androidTest/java/androidx/media3/muxer/Mp4MuxerEndToEndTest.java

@@ -46,7 +46,7 @@ import org.junit.runners.Parameterized.Parameters;
 /** End to end instrumentation tests for {@link Mp4Muxer}. */
 @RunWith(Parameterized.class)
 public class Mp4MuxerEndToEndTest {
-  private static final String H264_MP4 = "sample.mp4";
+  private static final String H264_MP4 = "sample_no_bframes.mp4";
   private static final String H265_HDR10_MP4 = "hdr10-720p.mp4";
   private static final String H265_WITH_METADATA_TRACK_MP4 = "h265_with_metadata_track.mp4";
   private static final String AV1_MP4 = "sample_av1.mp4";

libraries/muxer/src/main/java/androidx/media3/muxer/Mp4Muxer.java

@@ -242,6 +242,8 @@ public final class Mp4Muxer {
    * <p>The samples are cached and are written in batches so the caller must not change/release the
    * {@link ByteBuffer} and the {@link BufferInfo} after calling this method.
    *
+   * <p>Note: Out of order B-frames are currently not supported.
+   *
    * @param trackToken The {@link TrackToken} for which this sample is being written.
    * @param byteBuffer The encoded sample.
    * @param bufferInfo The {@link BufferInfo} related to this sample.

libraries/muxer/src/main/java/androidx/media3/muxer/Mp4Writer.java

@@ -15,8 +15,11 @@
  */
 package androidx.media3.muxer;
 
+import static androidx.media3.common.util.Assertions.checkArgument;
+
 import android.media.MediaCodec;
 import android.media.MediaCodec.BufferInfo;
+import androidx.media3.common.C;
 import androidx.media3.common.Format;
 import androidx.media3.common.MimeTypes;
 import androidx.media3.muxer.Mp4Muxer.TrackToken;
@@ -74,9 +77,10 @@ import java.util.List;
     public final List<Integer> writtenChunkSampleCounts;
     public final Deque<BufferInfo> pendingSamplesBufferInfo;
     public final Deque<ByteBuffer> pendingSamplesByteBuffer;
-
     public boolean hadKeyframe = false;
 
+    private long lastSamplePresentationTimeUs;
+
     /** Creates an instance with {@code sortKey} set to 1. */
     public Track(Format format) {
       this(format, /* sortKey= */ 1);
@@ -96,9 +100,13 @@ import java.util.List;
       writtenChunkSampleCounts = new ArrayList<>();
       pendingSamplesBufferInfo = new ArrayDeque<>();
       pendingSamplesByteBuffer = new ArrayDeque<>();
+      lastSamplePresentationTimeUs = C.TIME_UNSET;
     }
 
     public void writeSampleData(ByteBuffer byteBuffer, BufferInfo bufferInfo) throws IOException {
+      checkArgument(
+          bufferInfo.presentationTimeUs > lastSamplePresentationTimeUs,
+          "Out of order B-frames are not supported");
       // TODO: b/279931840 - Confirm whether muxer should throw when writing empty samples.
       //  Skip empty samples.
       if (bufferInfo.size == 0 || byteBuffer.remaining() == 0) {
@@ -109,12 +117,14 @@ import java.util.List;
         hadKeyframe = true;
       }
 
+      // The video track must start with a key frame.
       if (!hadKeyframe && MimeTypes.isVideo(format.sampleMimeType)) {
         return;
       }
 
       pendingSamplesBufferInfo.addLast(bufferInfo);
       pendingSamplesByteBuffer.addLast(byteBuffer);
+      lastSamplePresentationTimeUs = bufferInfo.presentationTimeUs;
     }
 
     @Override

libraries/muxer/src/test/java/androidx/media3/muxer/Mp4MuxerEndToEndTest.java

@@ -17,6 +17,7 @@ package androidx.media3.muxer;
 
 import static androidx.media3.muxer.MuxerTestUtil.FAKE_VIDEO_FORMAT;
 import static com.google.common.truth.Truth.assertThat;
+import static org.junit.Assert.assertThrows;
 
 import android.media.MediaCodec.BufferInfo;
 import android.util.Pair;
@@ -62,12 +63,10 @@ public class Mp4MuxerEndToEndTest {
     String outputFilePath = temporaryFolder.newFile().getPath();
     Mp4Muxer mp4Muxer = new Mp4Muxer.Builder(new FileOutputStream(outputFilePath)).build();
     mp4Muxer.setModificationTime(/* timestampMs= */ 500_000_000L);
-
     Pair<ByteBuffer, BufferInfo> track1Sample1 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 100L);
     Pair<ByteBuffer, BufferInfo> track1Sample2 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 200L);
-
     Pair<ByteBuffer, BufferInfo> track2Sample1 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 100L);
     Pair<ByteBuffer, BufferInfo> track2Sample2 =
@@ -104,12 +103,10 @@ public class Mp4MuxerEndToEndTest {
     String outputFilePath = temporaryFolder.newFile().getPath();
     Mp4Muxer mp4Muxer = new Mp4Muxer.Builder(new FileOutputStream(outputFilePath)).build();
     mp4Muxer.setModificationTime(/* timestampMs= */ 500_000_000L);
-
     Pair<ByteBuffer, BufferInfo> track1Sample1 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 0L);
     Pair<ByteBuffer, BufferInfo> track1Sample2 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 100L);
-
     Pair<ByteBuffer, BufferInfo> track2Sample1 =
         MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 100L);
     Pair<ByteBuffer, BufferInfo> track2Sample2 =
@@ -136,4 +133,27 @@ public class Mp4MuxerEndToEndTest {
         fakeExtractorOutput,
         MuxerTestUtil.getExpectedDumpFilePath("mp4_with_different_tracks_offset.mp4"));
   }
+
+  @Test
+  public void writeSampleData_withOutOfOrderSampleTimestamps_throws() throws IOException {
+    String outputFilePath = temporaryFolder.newFile().getPath();
+    Mp4Muxer mp4Muxer = new Mp4Muxer.Builder(new FileOutputStream(outputFilePath)).build();
+    Pair<ByteBuffer, BufferInfo> track1Sample1 =
+        MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 0L);
+    Pair<ByteBuffer, BufferInfo> track1Sample2 =
+        MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 2000L);
+    Pair<ByteBuffer, BufferInfo> track1Sample3 =
+        MuxerTestUtil.getFakeSampleAndSampleInfo(/* presentationTimeUs= */ 1000L);
+    try {
+      TrackToken track1 = mp4Muxer.addTrack(/* sortKey= */ 0, FAKE_VIDEO_FORMAT);
+      mp4Muxer.writeSampleData(track1, track1Sample1.first, track1Sample1.second);
+      mp4Muxer.writeSampleData(track1, track1Sample2.first, track1Sample2.second);
+
+      assertThrows(
+          IllegalArgumentException.class,
+          () -> mp4Muxer.writeSampleData(track1, track1Sample3.first, track1Sample3.second));
+    } finally {
+      mp4Muxer.close();
+    }
+  }
 }

libraries/test_data/src/test/assets/muxerdumps/sample_no_bframes.mp4.dump

@@ -7,149 +7,10 @@ seekMap:
   getPosition(1065600) = [[timeUs=0, position=44]]
 numberOfTracks = 2
 track 0:
-  total output bytes = 89876
-  sample count = 30
-  format 0:
-    id = 1
-    sampleMimeType = video/avc
-    codecs = avc1.64001F
-    maxInputSize = 36722
-    width = 1080
-    height = 720
-    frameRate = 32.113037
-    colorInfo:
-      lumaBitdepth = 8
-      chromaBitdepth = 8
-    metadata = entries=[Mp4Timestamp: creation time=2083344800, modification time=2083344800, timescale=10000]
-    initializationData:
-      data = length 29, hash 4746B5D9
-      data = length 10, hash 7A0D0F2B
-  sample 0:
-    time = 0
-    flags = 1
-    data = length 36692, hash D216076E
-  sample 1:
-    time = 66722
-    flags = 0
-    data = length 5312, hash D45D3CA0
-  sample 2:
-    time = 33355
-    flags = 0
-    data = length 599, hash 1BE7812D
-  sample 3:
-    time = 200200
-    flags = 0
-    data = length 7735, hash 4490F110
-  sample 4:
-    time = 133455
-    flags = 0
-    data = length 987, hash 560B5036
-  sample 5:
-    time = 100100
-    flags = 0
-    data = length 673, hash ED7CD8C7
-  sample 6:
-    time = 166822
-    flags = 0
-    data = length 523, hash 3020DF50
-  sample 7:
-    time = 333655
-    flags = 0
-    data = length 6061, hash 736C72B2
-  sample 8:
-    time = 266922
-    flags = 0
-    data = length 992, hash FE132F23
-  sample 9:
-    time = 233555
-    flags = 0
-    data = length 623, hash 5B2C1816
-  sample 10:
-    time = 300300
-    flags = 0
-    data = length 421, hash 742E69C1
-  sample 11:
-    time = 433755
-    flags = 0
-    data = length 4899, hash F72F86A1
-  sample 12:
-    time = 400400
-    flags = 0
-    data = length 568, hash 519A8E50
-  sample 13:
-    time = 367022
-    flags = 0
-    data = length 620, hash 3990AA39
-  sample 14:
-    time = 567222
-    flags = 0
-    data = length 5450, hash F06EC4AA
-  sample 15:
-    time = 500500
-    flags = 0
-    data = length 1051, hash 92DFA63A
-  sample 16:
-    time = 467122
-    flags = 0
-    data = length 874, hash 69587FB4
-  sample 17:
-    time = 533855
-    flags = 0
-    data = length 781, hash 36BE495B
-  sample 18:
-    time = 700700
-    flags = 0
-    data = length 4725, hash AC0C8CD3
-  sample 19:
-    time = 633955
-    flags = 0
-    data = length 1022, hash 5D8BFF34
-  sample 20:
-    time = 600600
-    flags = 0
-    data = length 790, hash 99413A99
-  sample 21:
-    time = 667322
-    flags = 0
-    data = length 610, hash 5E129290
-  sample 22:
-    time = 834155
-    flags = 0
-    data = length 2751, hash 769974CB
-  sample 23:
-    time = 767422
-    flags = 0
-    data = length 745, hash B78A477A
-  sample 24:
-    time = 734055
-    flags = 0
-    data = length 621, hash CF741E7A
-  sample 25:
-    time = 800800
-    flags = 0
-    data = length 505, hash 1DB4894E
-  sample 26:
-    time = 967622
-    flags = 0
-    data = length 1268, hash C15348DC
-  sample 27:
-    time = 900900
-    flags = 0
-    data = length 880, hash C2DE85D0
-  sample 28:
-    time = 867522
-    flags = 0
-    data = length 530, hash C98BC6A8
-  sample 29:
-    time = 934255
-    flags = 536870912
-    data = length 568, hash 4FE5C8EA
-track 1:
   total output bytes = 9529
   sample count = 45
   format 0:
-    peakBitrate = 200000
+    id = 1
-    id = 2
     sampleMimeType = audio/mp4a-latm
     codecs = mp4a.40.2
     maxInputSize = 294
@@ -339,4 +200,145 @@ track 1:
     time = 1065666
     flags = 536870913
     data = length 6, hash 31B22286
+track 1:
+  total output bytes = 301392
+  sample count = 30
+  format 0:
+    id = 2
+    sampleMimeType = video/avc
+    codecs = avc1.640034
+    maxInputSize = 22910
+    width = 1080
+    height = 720
+    frameRate = 31.004547
+    colorInfo:
+      colorSpace = 1
+      colorRange = 2
+      colorTransfer = 3
+      lumaBitdepth = 8
+      chromaBitdepth = 8
+    metadata = entries=[Mp4Timestamp: creation time=2083344800, modification time=2083344800, timescale=10000]
+    initializationData:
+      data = length 23, hash 33E412EE
+      data = length 9, hash FBAFBC0C
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 7744, hash DDF91733
+  sample 1:
+    time = 33355
+    flags = 0
+    data = length 1168, hash 89E48A20
+  sample 2:
+    time = 66722
+    flags = 0
+    data = length 960, hash D4AD9EF0
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 976, hash CD49C23C
+  sample 4:
+    time = 133455
+    flags = 0
+    data = length 1360, hash 337B78A9
+  sample 5:
+    time = 166822
+    flags = 0
+    data = length 2288, hash 5D5FD1C8
+  sample 6:
+    time = 200200
+    flags = 0
+    data = length 3856, hash 3D7DCD46
+  sample 7:
+    time = 233555
+    flags = 0
+    data = length 4048, hash 47C78814
+  sample 8:
+    time = 266922
+    flags = 0
+    data = length 6144, hash 8FD9AD7D
+  sample 9:
+    time = 300300
+    flags = 0
+    data = length 7632, hash 4245F848
+  sample 10:
+    time = 333655
+    flags = 0
+    data = length 9792, hash B2B9AB4B
+  sample 11:
+    time = 367022
+    flags = 0
+    data = length 14496, hash E0F2E0BA
+  sample 12:
+    time = 400400
+    flags = 0
+    data = length 17664, hash 3E3189E
+  sample 13:
+    time = 433755
+    flags = 0
+    data = length 5712, hash CA808ECF
+  sample 14:
+    time = 467122
+    flags = 0
+    data = length 9776, hash C875D1AA
+  sample 15:
+    time = 500500
+    flags = 0
+    data = length 17712, hash 69AE17D4
+  sample 16:
+    time = 533855
+    flags = 0
+    data = length 11440, hash 7370E78C
+  sample 17:
+    time = 567222
+    flags = 0
+    data = length 8544, hash 5A581986
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 19904, hash 98AB5C44
+  sample 19:
+    time = 633955
+    flags = 0
+    data = length 14352, hash 74B754E3
+  sample 20:
+    time = 667322
+    flags = 0
+    data = length 9568, hash 369746A6
+  sample 21:
+    time = 700700
+    flags = 0
+    data = length 12192, hash E0F8A71A
+  sample 22:
+    time = 734055
+    flags = 0
+    data = length 22880, hash 75E833BA
+  sample 23:
+    time = 767422
+    flags = 0
+    data = length 16832, hash E8BFCFE3
+  sample 24:
+    time = 800800
+    flags = 0
+    data = length 5120, hash E04AEF94
+  sample 25:
+    time = 834155
+    flags = 0
+    data = length 9888, hash 1166103E
+  sample 26:
+    time = 867522
+    flags = 0
+    data = length 17024, hash F9A96740
+  sample 27:
+    time = 900900
+    flags = 0
+    data = length 20864, hash DF9E88B8
+  sample 28:
+    time = 934255
+    flags = 0
+    data = length 7216, hash BE22BE2F
+  sample 29:
+    time = 967622
+    flags = 536870912
+    data = length 14240, hash E190BF31
 tracksEnded = true

libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4/with_capture_fps.dump

@@ -7,149 +7,150 @@ seekMap:
   getPosition(1065600) = [[timeUs=0, position=44]]
 numberOfTracks = 2
 track 0:
-  total output bytes = 89876
+  total output bytes = 301392
   sample count = 30
   format 0:
     id = 1
     sampleMimeType = video/avc
-    codecs = avc1.64001F
+    codecs = avc1.640034
-    maxInputSize = 36722
+    maxInputSize = 22910
     width = 1080
     height = 720
-    frameRate = 32.113037
+    frameRate = 31.004547
     colorInfo:
+      colorSpace = 1
       colorRange = 2
+      colorTransfer = 3
       lumaBitdepth = 8
       chromaBitdepth = 8
-    metadata = entries=[mdta: key=com.android.capture.fps, value=60.0, xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, mdta: key=com.android.capture.fps, value=60.0, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
-      data = length 29, hash 4746B5D9
+      data = length 23, hash 33E412EE
-      data = length 10, hash 7A0D0F2B
+      data = length 9, hash FBAFBC0C
   sample 0:
     time = 0
     flags = 1
-    data = length 36692, hash D216076E
+    data = length 7744, hash DDF91733
   sample 1:
-    time = 66722
-    flags = 0
-    data = length 5312, hash D45D3CA0
-  sample 2:
     time = 33355
     flags = 0
-    data = length 599, hash 1BE7812D
+    data = length 1168, hash 89E48A20
-  sample 3:
+  sample 2:
-    time = 200200
+    time = 66722
     flags = 0
-    data = length 7735, hash 4490F110
+    data = length 960, hash D4AD9EF0
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 976, hash CD49C23C
   sample 4:
     time = 133455
     flags = 0
-    data = length 987, hash 560B5036
+    data = length 1360, hash 337B78A9
   sample 5:
-    time = 100100
-    flags = 0
-    data = length 673, hash ED7CD8C7
-  sample 6:
     time = 166822
     flags = 0
-    data = length 523, hash 3020DF50
+    data = length 2288, hash 5D5FD1C8
-  sample 7:
+  sample 6:
-    time = 333655
+    time = 200200
     flags = 0
-    data = length 6061, hash 736C72B2
+    data = length 3856, hash 3D7DCD46
+  sample 7:
+    time = 233555
+    flags = 0
+    data = length 4048, hash 47C78814
   sample 8:
     time = 266922
     flags = 0
-    data = length 992, hash FE132F23
+    data = length 6144, hash 8FD9AD7D
   sample 9:
-    time = 233555
-    flags = 0
-    data = length 623, hash 5B2C1816
-  sample 10:
     time = 300300
     flags = 0
-    data = length 421, hash 742E69C1
+    data = length 7632, hash 4245F848
-  sample 11:
+  sample 10:
-    time = 433755
+    time = 333655
     flags = 0
-    data = length 4899, hash F72F86A1
+    data = length 9792, hash B2B9AB4B
+  sample 11:
+    time = 367022
+    flags = 0
+    data = length 14496, hash E0F2E0BA
   sample 12:
     time = 400400
     flags = 0
-    data = length 568, hash 519A8E50
+    data = length 17664, hash 3E3189E
   sample 13:
-    time = 367022
+    time = 433755
     flags = 0
-    data = length 620, hash 3990AA39
+    data = length 5712, hash CA808ECF
   sample 14:
-    time = 567222
+    time = 467122
     flags = 0
-    data = length 5450, hash F06EC4AA
+    data = length 9776, hash C875D1AA
   sample 15:
     time = 500500
     flags = 0
-    data = length 1051, hash 92DFA63A
+    data = length 17712, hash 69AE17D4
   sample 16:
-    time = 467122
-    flags = 0
-    data = length 874, hash 69587FB4
-  sample 17:
     time = 533855
     flags = 0
-    data = length 781, hash 36BE495B
+    data = length 11440, hash 7370E78C
-  sample 18:
+  sample 17:
-    time = 700700
+    time = 567222
     flags = 0
-    data = length 4725, hash AC0C8CD3
+    data = length 8544, hash 5A581986
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 19904, hash 98AB5C44
   sample 19:
     time = 633955
     flags = 0
-    data = length 1022, hash 5D8BFF34
+    data = length 14352, hash 74B754E3
   sample 20:
-    time = 600600
-    flags = 0
-    data = length 790, hash 99413A99
-  sample 21:
     time = 667322
     flags = 0
-    data = length 610, hash 5E129290
+    data = length 9568, hash 369746A6
-  sample 22:
+  sample 21:
-    time = 834155
+    time = 700700
     flags = 0
-    data = length 2751, hash 769974CB
+    data = length 12192, hash E0F8A71A
+  sample 22:
+    time = 734055
+    flags = 0
+    data = length 22880, hash 75E833BA
   sample 23:
     time = 767422
     flags = 0
-    data = length 745, hash B78A477A
+    data = length 16832, hash E8BFCFE3
   sample 24:
-    time = 734055
-    flags = 0
-    data = length 621, hash CF741E7A
-  sample 25:
     time = 800800
     flags = 0
-    data = length 505, hash 1DB4894E
+    data = length 5120, hash E04AEF94
-  sample 26:
+  sample 25:
-    time = 967622
+    time = 834155
     flags = 0
-    data = length 1268, hash C15348DC
+    data = length 9888, hash 1166103E
+  sample 26:
+    time = 867522
+    flags = 0
+    data = length 17024, hash F9A96740
   sample 27:
     time = 900900
     flags = 0
-    data = length 880, hash C2DE85D0
+    data = length 20864, hash DF9E88B8
   sample 28:
-    time = 867522
-    flags = 0
-    data = length 530, hash C98BC6A8
-  sample 29:
     time = 934255
+    flags = 0
+    data = length 7216, hash BE22BE2F
+  sample 29:
+    time = 967622
     flags = 536870912
-    data = length 568, hash 4FE5C8EA
+    data = length 14240, hash E190BF31
 track 1:
   total output bytes = 9529
   sample count = 45
   format 0:
-    peakBitrate = 200000
     id = 2
     sampleMimeType = audio/mp4a-latm
     codecs = mp4a.40.2
@@ -157,7 +158,7 @@ track 1:
     channelCount = 1
     sampleRate = 44100
     language = und
-    metadata = entries=[xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
       data = length 2, hash 5F7
   sample 0:

libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4/with_creation_time.dump

@@ -7,149 +7,150 @@ seekMap:
   getPosition(1065600) = [[timeUs=0, position=44]]
 numberOfTracks = 2
 track 0:
-  total output bytes = 89876
+  total output bytes = 301392
   sample count = 30
   format 0:
     id = 1
     sampleMimeType = video/avc
-    codecs = avc1.64001F
+    codecs = avc1.640034
-    maxInputSize = 36722
+    maxInputSize = 22910
     width = 1080
     height = 720
-    frameRate = 32.113037
+    frameRate = 31.004547
     colorInfo:
+      colorSpace = 1
       colorRange = 2
+      colorTransfer = 3
       lumaBitdepth = 8
       chromaBitdepth = 8
-    metadata = entries=[xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=2000000000, modification time=2000000000, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=2000000000, modification time=2000000000, timescale=10000]
     initializationData:
-      data = length 29, hash 4746B5D9
+      data = length 23, hash 33E412EE
-      data = length 10, hash 7A0D0F2B
+      data = length 9, hash FBAFBC0C
   sample 0:
     time = 0
     flags = 1
-    data = length 36692, hash D216076E
+    data = length 7744, hash DDF91733
   sample 1:
-    time = 66722
-    flags = 0
-    data = length 5312, hash D45D3CA0
-  sample 2:
     time = 33355
     flags = 0
-    data = length 599, hash 1BE7812D
+    data = length 1168, hash 89E48A20
-  sample 3:
+  sample 2:
-    time = 200200
+    time = 66722
     flags = 0
-    data = length 7735, hash 4490F110
+    data = length 960, hash D4AD9EF0
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 976, hash CD49C23C
   sample 4:
     time = 133455
     flags = 0
-    data = length 987, hash 560B5036
+    data = length 1360, hash 337B78A9
   sample 5:
-    time = 100100
-    flags = 0
-    data = length 673, hash ED7CD8C7
-  sample 6:
     time = 166822
     flags = 0
-    data = length 523, hash 3020DF50
+    data = length 2288, hash 5D5FD1C8
-  sample 7:
+  sample 6:
-    time = 333655
+    time = 200200
     flags = 0
-    data = length 6061, hash 736C72B2
+    data = length 3856, hash 3D7DCD46
+  sample 7:
+    time = 233555
+    flags = 0
+    data = length 4048, hash 47C78814
   sample 8:
     time = 266922
     flags = 0
-    data = length 992, hash FE132F23
+    data = length 6144, hash 8FD9AD7D
   sample 9:
-    time = 233555
-    flags = 0
-    data = length 623, hash 5B2C1816
-  sample 10:
     time = 300300
     flags = 0
-    data = length 421, hash 742E69C1
+    data = length 7632, hash 4245F848
-  sample 11:
+  sample 10:
-    time = 433755
+    time = 333655
     flags = 0
-    data = length 4899, hash F72F86A1
+    data = length 9792, hash B2B9AB4B
+  sample 11:
+    time = 367022
+    flags = 0
+    data = length 14496, hash E0F2E0BA
   sample 12:
     time = 400400
     flags = 0
-    data = length 568, hash 519A8E50
+    data = length 17664, hash 3E3189E
   sample 13:
-    time = 367022
+    time = 433755
     flags = 0
-    data = length 620, hash 3990AA39
+    data = length 5712, hash CA808ECF
   sample 14:
-    time = 567222
+    time = 467122
     flags = 0
-    data = length 5450, hash F06EC4AA
+    data = length 9776, hash C875D1AA
   sample 15:
     time = 500500
     flags = 0
-    data = length 1051, hash 92DFA63A
+    data = length 17712, hash 69AE17D4
   sample 16:
-    time = 467122
-    flags = 0
-    data = length 874, hash 69587FB4
-  sample 17:
     time = 533855
     flags = 0
-    data = length 781, hash 36BE495B
+    data = length 11440, hash 7370E78C
-  sample 18:
+  sample 17:
-    time = 700700
+    time = 567222
     flags = 0
-    data = length 4725, hash AC0C8CD3
+    data = length 8544, hash 5A581986
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 19904, hash 98AB5C44
   sample 19:
     time = 633955
     flags = 0
-    data = length 1022, hash 5D8BFF34
+    data = length 14352, hash 74B754E3
   sample 20:
-    time = 600600
-    flags = 0
-    data = length 790, hash 99413A99
-  sample 21:
     time = 667322
     flags = 0
-    data = length 610, hash 5E129290
+    data = length 9568, hash 369746A6
-  sample 22:
+  sample 21:
-    time = 834155
+    time = 700700
     flags = 0
-    data = length 2751, hash 769974CB
+    data = length 12192, hash E0F8A71A
+  sample 22:
+    time = 734055
+    flags = 0
+    data = length 22880, hash 75E833BA
   sample 23:
     time = 767422
     flags = 0
-    data = length 745, hash B78A477A
+    data = length 16832, hash E8BFCFE3
   sample 24:
-    time = 734055
-    flags = 0
-    data = length 621, hash CF741E7A
-  sample 25:
     time = 800800
     flags = 0
-    data = length 505, hash 1DB4894E
+    data = length 5120, hash E04AEF94
-  sample 26:
+  sample 25:
-    time = 967622
+    time = 834155
     flags = 0
-    data = length 1268, hash C15348DC
+    data = length 9888, hash 1166103E
+  sample 26:
+    time = 867522
+    flags = 0
+    data = length 17024, hash F9A96740
   sample 27:
     time = 900900
     flags = 0
-    data = length 880, hash C2DE85D0
+    data = length 20864, hash DF9E88B8
   sample 28:
-    time = 867522
-    flags = 0
-    data = length 530, hash C98BC6A8
-  sample 29:
     time = 934255
+    flags = 0
+    data = length 7216, hash BE22BE2F
+  sample 29:
+    time = 967622
     flags = 536870912
-    data = length 568, hash 4FE5C8EA
+    data = length 14240, hash E190BF31
 track 1:
   total output bytes = 9529
   sample count = 45
   format 0:
-    peakBitrate = 200000
     id = 2
     sampleMimeType = audio/mp4a-latm
     codecs = mp4a.40.2
@@ -157,7 +158,7 @@ track 1:
     channelCount = 1
     sampleRate = 44100
     language = und
-    metadata = entries=[xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=2000000000, modification time=2000000000, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=2000000000, modification time=2000000000, timescale=10000]
     initializationData:
       data = length 2, hash 5F7
   sample 0:

libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4/with_custom_metadata.dump

@@ -7,149 +7,150 @@ seekMap:
   getPosition(1065600) = [[timeUs=0, position=44]]
 numberOfTracks = 2
 track 0:
-  total output bytes = 89876
+  total output bytes = 301392
   sample count = 30
   format 0:
     id = 1
     sampleMimeType = video/avc
-    codecs = avc1.64001F
+    codecs = avc1.640034
-    maxInputSize = 36722
+    maxInputSize = 22910
     width = 1080
     height = 720
-    frameRate = 32.113037
+    frameRate = 31.004547
     colorInfo:
+      colorSpace = 1
       colorRange = 2
+      colorTransfer = 3
       lumaBitdepth = 8
       chromaBitdepth = 8
-    metadata = entries=[mdta: key=StringKey, value=StringValue, mdta: key=FloatKey, value=600.0, xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, mdta: key=StringKey, value=StringValue, mdta: key=FloatKey, value=600.0, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
-      data = length 29, hash 4746B5D9
+      data = length 23, hash 33E412EE
-      data = length 10, hash 7A0D0F2B
+      data = length 9, hash FBAFBC0C
   sample 0:
     time = 0
     flags = 1
-    data = length 36692, hash D216076E
+    data = length 7744, hash DDF91733
   sample 1:
-    time = 66722
-    flags = 0
-    data = length 5312, hash D45D3CA0
-  sample 2:
     time = 33355
     flags = 0
-    data = length 599, hash 1BE7812D
+    data = length 1168, hash 89E48A20
-  sample 3:
+  sample 2:
-    time = 200200
+    time = 66722
     flags = 0
-    data = length 7735, hash 4490F110
+    data = length 960, hash D4AD9EF0
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 976, hash CD49C23C
   sample 4:
     time = 133455
     flags = 0
-    data = length 987, hash 560B5036
+    data = length 1360, hash 337B78A9
   sample 5:
-    time = 100100
-    flags = 0
-    data = length 673, hash ED7CD8C7
-  sample 6:
     time = 166822
     flags = 0
-    data = length 523, hash 3020DF50
+    data = length 2288, hash 5D5FD1C8
-  sample 7:
+  sample 6:
-    time = 333655
+    time = 200200
     flags = 0
-    data = length 6061, hash 736C72B2
+    data = length 3856, hash 3D7DCD46
+  sample 7:
+    time = 233555
+    flags = 0
+    data = length 4048, hash 47C78814
   sample 8:
     time = 266922
     flags = 0
-    data = length 992, hash FE132F23
+    data = length 6144, hash 8FD9AD7D
   sample 9:
-    time = 233555
-    flags = 0
-    data = length 623, hash 5B2C1816
-  sample 10:
     time = 300300
     flags = 0
-    data = length 421, hash 742E69C1
+    data = length 7632, hash 4245F848
-  sample 11:
+  sample 10:
-    time = 433755
+    time = 333655
     flags = 0
-    data = length 4899, hash F72F86A1
+    data = length 9792, hash B2B9AB4B
+  sample 11:
+    time = 367022
+    flags = 0
+    data = length 14496, hash E0F2E0BA
   sample 12:
     time = 400400
     flags = 0
-    data = length 568, hash 519A8E50
+    data = length 17664, hash 3E3189E
   sample 13:
-    time = 367022
+    time = 433755
     flags = 0
-    data = length 620, hash 3990AA39
+    data = length 5712, hash CA808ECF
   sample 14:
-    time = 567222
+    time = 467122
     flags = 0
-    data = length 5450, hash F06EC4AA
+    data = length 9776, hash C875D1AA
   sample 15:
     time = 500500
     flags = 0
-    data = length 1051, hash 92DFA63A
+    data = length 17712, hash 69AE17D4
   sample 16:
-    time = 467122
-    flags = 0
-    data = length 874, hash 69587FB4
-  sample 17:
     time = 533855
     flags = 0
-    data = length 781, hash 36BE495B
+    data = length 11440, hash 7370E78C
-  sample 18:
+  sample 17:
-    time = 700700
+    time = 567222
     flags = 0
-    data = length 4725, hash AC0C8CD3
+    data = length 8544, hash 5A581986
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 19904, hash 98AB5C44
   sample 19:
     time = 633955
     flags = 0
-    data = length 1022, hash 5D8BFF34
+    data = length 14352, hash 74B754E3
   sample 20:
-    time = 600600
-    flags = 0
-    data = length 790, hash 99413A99
-  sample 21:
     time = 667322
     flags = 0
-    data = length 610, hash 5E129290
+    data = length 9568, hash 369746A6
-  sample 22:
+  sample 21:
-    time = 834155
+    time = 700700
     flags = 0
-    data = length 2751, hash 769974CB
+    data = length 12192, hash E0F8A71A
+  sample 22:
+    time = 734055
+    flags = 0
+    data = length 22880, hash 75E833BA
   sample 23:
     time = 767422
     flags = 0
-    data = length 745, hash B78A477A
+    data = length 16832, hash E8BFCFE3
   sample 24:
-    time = 734055
-    flags = 0
-    data = length 621, hash CF741E7A
-  sample 25:
     time = 800800
     flags = 0
-    data = length 505, hash 1DB4894E
+    data = length 5120, hash E04AEF94
-  sample 26:
+  sample 25:
-    time = 967622
+    time = 834155
     flags = 0
-    data = length 1268, hash C15348DC
+    data = length 9888, hash 1166103E
+  sample 26:
+    time = 867522
+    flags = 0
+    data = length 17024, hash F9A96740
   sample 27:
     time = 900900
     flags = 0
-    data = length 880, hash C2DE85D0
+    data = length 20864, hash DF9E88B8
   sample 28:
-    time = 867522
-    flags = 0
-    data = length 530, hash C98BC6A8
-  sample 29:
     time = 934255
+    flags = 0
+    data = length 7216, hash BE22BE2F
+  sample 29:
+    time = 967622
     flags = 536870912
-    data = length 568, hash 4FE5C8EA
+    data = length 14240, hash E190BF31
 track 1:
   total output bytes = 9529
   sample count = 45
   format 0:
-    peakBitrate = 200000
     id = 2
     sampleMimeType = audio/mp4a-latm
     codecs = mp4a.40.2
@@ -157,7 +158,7 @@ track 1:
     channelCount = 1
     sampleRate = 44100
     language = und
-    metadata = entries=[mdta: key=StringKey, value=StringValue, mdta: key=FloatKey, value=600.0, xyz: latitude=40.68, longitude=-74.5, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, mdta: key=StringKey, value=StringValue, mdta: key=FloatKey, value=600.0, xyz: latitude=40.68, longitude=-74.4999, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
       data = length 2, hash 5F7
   sample 0:

libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4/with_location_metadata.dump

@@ -7,149 +7,150 @@ seekMap:
   getPosition(1065600) = [[timeUs=0, position=44]]
 numberOfTracks = 2
 track 0:
-  total output bytes = 89876
+  total output bytes = 301392
   sample count = 30
   format 0:
     id = 1
     sampleMimeType = video/avc
-    codecs = avc1.64001F
+    codecs = avc1.640034
-    maxInputSize = 36722
+    maxInputSize = 22910
     width = 1080
     height = 720
-    frameRate = 32.113037
+    frameRate = 31.004547
     colorInfo:
+      colorSpace = 1
       colorRange = 2
+      colorTransfer = 3
       lumaBitdepth = 8
       chromaBitdepth = 8
-    metadata = entries=[xyz: latitude=45.0, longitude=-90.0, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, xyz: latitude=45.0, longitude=-90.0, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
-      data = length 29, hash 4746B5D9
+      data = length 23, hash 33E412EE
-      data = length 10, hash 7A0D0F2B
+      data = length 9, hash FBAFBC0C
   sample 0:
     time = 0
     flags = 1
-    data = length 36692, hash D216076E
+    data = length 7744, hash DDF91733
   sample 1:
-    time = 66722
-    flags = 0
-    data = length 5312, hash D45D3CA0
-  sample 2:
     time = 33355
     flags = 0
-    data = length 599, hash 1BE7812D
+    data = length 1168, hash 89E48A20
-  sample 3:
+  sample 2:
-    time = 200200
+    time = 66722
     flags = 0
-    data = length 7735, hash 4490F110
+    data = length 960, hash D4AD9EF0
+  sample 3:
+    time = 100100
+    flags = 0
+    data = length 976, hash CD49C23C
   sample 4:
     time = 133455
     flags = 0
-    data = length 987, hash 560B5036
+    data = length 1360, hash 337B78A9
   sample 5:
-    time = 100100
-    flags = 0
-    data = length 673, hash ED7CD8C7
-  sample 6:
     time = 166822
     flags = 0
-    data = length 523, hash 3020DF50
+    data = length 2288, hash 5D5FD1C8
-  sample 7:
+  sample 6:
-    time = 333655
+    time = 200200
     flags = 0
-    data = length 6061, hash 736C72B2
+    data = length 3856, hash 3D7DCD46
+  sample 7:
+    time = 233555
+    flags = 0
+    data = length 4048, hash 47C78814
   sample 8:
     time = 266922
     flags = 0
-    data = length 992, hash FE132F23
+    data = length 6144, hash 8FD9AD7D
   sample 9:
-    time = 233555
-    flags = 0
-    data = length 623, hash 5B2C1816
-  sample 10:
     time = 300300
     flags = 0
-    data = length 421, hash 742E69C1
+    data = length 7632, hash 4245F848
-  sample 11:
+  sample 10:
-    time = 433755
+    time = 333655
     flags = 0
-    data = length 4899, hash F72F86A1
+    data = length 9792, hash B2B9AB4B
+  sample 11:
+    time = 367022
+    flags = 0
+    data = length 14496, hash E0F2E0BA
   sample 12:
     time = 400400
     flags = 0
-    data = length 568, hash 519A8E50
+    data = length 17664, hash 3E3189E
   sample 13:
-    time = 367022
+    time = 433755
     flags = 0
-    data = length 620, hash 3990AA39
+    data = length 5712, hash CA808ECF
   sample 14:
-    time = 567222
+    time = 467122
     flags = 0
-    data = length 5450, hash F06EC4AA
+    data = length 9776, hash C875D1AA
   sample 15:
     time = 500500
     flags = 0
-    data = length 1051, hash 92DFA63A
+    data = length 17712, hash 69AE17D4
   sample 16:
-    time = 467122
-    flags = 0
-    data = length 874, hash 69587FB4
-  sample 17:
     time = 533855
     flags = 0
-    data = length 781, hash 36BE495B
+    data = length 11440, hash 7370E78C
-  sample 18:
+  sample 17:
-    time = 700700
+    time = 567222
     flags = 0
-    data = length 4725, hash AC0C8CD3
+    data = length 8544, hash 5A581986
+  sample 18:
+    time = 600600
+    flags = 0
+    data = length 19904, hash 98AB5C44
   sample 19:
     time = 633955
     flags = 0
-    data = length 1022, hash 5D8BFF34
+    data = length 14352, hash 74B754E3
   sample 20:
-    time = 600600
-    flags = 0
-    data = length 790, hash 99413A99
-  sample 21:
     time = 667322
     flags = 0
-    data = length 610, hash 5E129290
+    data = length 9568, hash 369746A6
-  sample 22:
+  sample 21:
-    time = 834155
+    time = 700700
     flags = 0
-    data = length 2751, hash 769974CB
+    data = length 12192, hash E0F8A71A
+  sample 22:
+    time = 734055
+    flags = 0
+    data = length 22880, hash 75E833BA
   sample 23:
     time = 767422
     flags = 0
-    data = length 745, hash B78A477A
+    data = length 16832, hash E8BFCFE3
   sample 24:
-    time = 734055
-    flags = 0
-    data = length 621, hash CF741E7A
-  sample 25:
     time = 800800
     flags = 0
-    data = length 505, hash 1DB4894E
+    data = length 5120, hash E04AEF94
-  sample 26:
+  sample 25:
-    time = 967622
+    time = 834155
     flags = 0
-    data = length 1268, hash C15348DC
+    data = length 9888, hash 1166103E
+  sample 26:
+    time = 867522
+    flags = 0
+    data = length 17024, hash F9A96740
   sample 27:
     time = 900900
     flags = 0
-    data = length 880, hash C2DE85D0
+    data = length 20864, hash DF9E88B8
   sample 28:
-    time = 867522
-    flags = 0
-    data = length 530, hash C98BC6A8
-  sample 29:
     time = 934255
+    flags = 0
+    data = length 7216, hash BE22BE2F
+  sample 29:
+    time = 967622
     flags = 536870912
-    data = length 568, hash 4FE5C8EA
+    data = length 14240, hash E190BF31
 track 1:
   total output bytes = 9529
   sample count = 45
   format 0:
-    peakBitrate = 200000
     id = 2
     sampleMimeType = audio/mp4a-latm
     codecs = mp4a.40.2
@@ -157,7 +158,7 @@ track 1:
     channelCount = 1
     sampleRate = 44100
     language = und
-    metadata = entries=[xyz: latitude=45.0, longitude=-90.0, Mp4Timestamp: creation time=3547558895, modification time=3547558895, timescale=10000]
+    metadata = entries=[mdta: key=com.android.version, value=13, xyz: latitude=45.0, longitude=-90.0, Mp4Timestamp: creation time=3785416745, modification time=3785416745, timescale=10000]
     initializationData:
       data = length 2, hash 5F7
   sample 0:

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

@@ -17,7 +17,6 @@ package androidx.media3.transformer;
 
 import static androidx.media3.common.util.Assertions.checkNotNull;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_FORMAT;
-import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_URI_STRING;
 import static org.junit.Assume.assumeTrue;
 
 import android.content.Context;
@@ -40,7 +39,7 @@ import org.junit.runners.Parameterized.Parameters;
 @RunWith(Parameterized.class)
 public class TransformerWithInAppMuxerEndToEndTest {
   private static final String MP4_FILE_ASSET_DIRECTORY = "asset:///media/mp4/";
-  private static final String H264_MP4 = "sample.mp4";
+  private static final String H264_MP4 = "sample_no_bframes.mp4";
   private static final String H265_MP4 = "h265_with_metadata_track.mp4";
 
   @Parameters(name = "{0}")
@@ -90,7 +89,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
     ChannelMixingAudioProcessor channelMixingAudioProcessor = new ChannelMixingAudioProcessor();
     channelMixingAudioProcessor.putChannelMixingMatrix(
         ChannelMixingMatrix.create(/* inputChannelCount= */ 1, /* outputChannelCount= */ 2));
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_ASSET_URI_STRING));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE_ASSET_DIRECTORY + H264_MP4));
     EditedMediaItem editedMediaItem =
         new EditedMediaItem.Builder(mediaItem)
             .setEffects(

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

@@ -16,7 +16,6 @@
 package androidx.media3.transformer;
 
 import static androidx.media3.common.util.Assertions.checkNotNull;
-import static androidx.media3.transformer.TestUtil.ASSET_URI_PREFIX;
 import static androidx.media3.transformer.TestUtil.FILE_AUDIO_VIDEO;
 import static com.google.common.truth.Truth.assertThat;
 
@@ -44,8 +43,7 @@ import org.junit.runner.RunWith;
 /** End-to-end test for {@link Transformer} with {@link InAppMuxer}. */
 @RunWith(AndroidJUnit4.class)
 public class TransformerWithInAppMuxerEndToEndTest {
-  private static final String XMP_SAMPLE_DATA = "media/xmp/sample_datetime_xmp.xmp";
+  private static final String MP4_FILE = "asset:///media/mp4/sample_no_bframes.mp4";
-
   @Rule public final TemporaryFolder outputDir = new TemporaryFolder();
 
   private final Context context = ApplicationProvider.getApplicationContext();
@@ -70,7 +68,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .setMuxerFactory(inAppMuxerFactory)
             .build();
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(ASSET_URI_PREFIX + FILE_AUDIO_VIDEO));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE));
 
     transformer.start(mediaItem, outputPath);
     TransformerTestRunner.runLooper(transformer);
@@ -89,7 +87,8 @@ public class TransformerWithInAppMuxerEndToEndTest {
 
   @Test
   public void transmux_withXmpData_completesSuccessfully() throws Exception {
-    byte[] xmpData = androidx.media3.test.utils.TestUtil.getByteArray(context, XMP_SAMPLE_DATA);
+    String xmpSampleData = "media/xmp/sample_datetime_xmp.xmp";
+    byte[] xmpData = androidx.media3.test.utils.TestUtil.getByteArray(context, xmpSampleData);
     Muxer.Factory inAppMuxerFactory =
         new InAppMuxer.Factory(
             DefaultMuxer.Factory.DEFAULT_MAX_DELAY_BETWEEN_SAMPLES_MS,
@@ -99,7 +98,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .setMuxerFactory(inAppMuxerFactory)
             .build();
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(ASSET_URI_PREFIX + FILE_AUDIO_VIDEO));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE));
 
     transformer.start(mediaItem, outputPath);
     ExportResult exportResult = TransformerTestRunner.runLooper(transformer);
@@ -127,7 +126,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .setMuxerFactory(inAppMuxerFactory)
             .build();
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(ASSET_URI_PREFIX + FILE_AUDIO_VIDEO));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE));
 
     transformer.start(mediaItem, outputPath);
     TransformerTestRunner.runLooper(transformer);
@@ -156,7 +155,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .setMuxerFactory(inAppMuxerFactory)
             .build();
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(ASSET_URI_PREFIX + FILE_AUDIO_VIDEO));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE));
 
     transformer.start(mediaItem, outputPath);
     TransformerTestRunner.runLooper(transformer);
@@ -201,7 +200,7 @@ public class TransformerWithInAppMuxerEndToEndTest {
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .setMuxerFactory(inAppMuxerFactory)
             .build();
-    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(ASSET_URI_PREFIX + FILE_AUDIO_VIDEO));
+    MediaItem mediaItem = MediaItem.fromUri(Uri.parse(MP4_FILE));
 
     transformer.start(mediaItem, outputPath);
     TransformerTestRunner.runLooper(transformer);