Transformer: Add support for transmuxing audio in trim optimization

PiperOrigin-RevId: 588711597

libraries/test_data/src/test/assets/transformerdumps/mp4/sample_with_increasing_timestamps_320w_240h.mp4/trimOptimizedClippedAtKeyFrame.dump

@@ -1,560 +0,0 @@
-format video:
-  id = 1
-  sampleMimeType = video/avc
-  codecs = avc1.42C015
-  maxInputSize = 14839
-  width = 320
-  height = 240
-  frameRate = 59.997425
-  colorInfo:
-    colorSpace = 2
-    colorRange = 1
-    colorTransfer = 3
-    lumaBitdepth = 8
-    chromaBitdepth = 8
-  metadata = entries=[TSSE: description=null: values=[Lavf58.76.100], Mp4Timestamp: creation time=0, modification time=0, timescale=1000]
-  initializationData:
-    data = length 31, hash 4B108214
-    data = length 9, hash FBA158BB
-container metadata = entries=[TSSE: description=null: values=[Lavf58.76.100], Mp4Timestamp: creation time=0, modification time=0, timescale=1000]
-sample:
-  trackType = video
-  dataHashCode = 983000500
-  size = 13539
-  isKeyFrame = true
-  presentationTimeUs = 0
-sample:
-  trackType = video
-  dataHashCode = -1834230781
-  size = 32
-  isKeyFrame = false
-  presentationTimeUs = 16666
-sample:
-  trackType = video
-  dataHashCode = 521720738
-  size = 1534
-  isKeyFrame = false
-  presentationTimeUs = 33333
-sample:
-  trackType = video
-  dataHashCode = 722836039
-  size = 123
-  isKeyFrame = false
-  presentationTimeUs = 50000
-sample:
-  trackType = video
-  dataHashCode = -1702585381
-  size = 2061
-  isKeyFrame = false
-  presentationTimeUs = 66666
-sample:
-  trackType = video
-  dataHashCode = -365856396
-  size = 147
-  isKeyFrame = false
-  presentationTimeUs = 83333
-sample:
-  trackType = video
-  dataHashCode = 1258185334
-  size = 2534
-  isKeyFrame = false
-  presentationTimeUs = 100000
-sample:
-  trackType = video
-  dataHashCode = -179623006
-  size = 87
-  isKeyFrame = false
-  presentationTimeUs = 116666
-sample:
-  trackType = video
-  dataHashCode = -541393824
-  size = 2762
-  isKeyFrame = false
-  presentationTimeUs = 133333
-sample:
-  trackType = video
-  dataHashCode = -1912932514
-  size = 57
-  isKeyFrame = false
-  presentationTimeUs = 150000
-sample:
-  trackType = video
-  dataHashCode = 485634444
-  size = 2833
-  isKeyFrame = false
-  presentationTimeUs = 166666
-sample:
-  trackType = video
-  dataHashCode = 570625802
-  size = 189
-  isKeyFrame = false
-  presentationTimeUs = 183333
-sample:
-  trackType = video
-  dataHashCode = 1819668957
-  size = 3153
-  isKeyFrame = false
-  presentationTimeUs = 200000
-sample:
-  trackType = video
-  dataHashCode = 1004398066
-  size = 104
-  isKeyFrame = false
-  presentationTimeUs = 216666
-sample:
-  trackType = video
-  dataHashCode = 2087741113
-  size = 2304
-  isKeyFrame = false
-  presentationTimeUs = 233333
-sample:
-  trackType = video
-  dataHashCode = -419782502
-  size = 222
-  isKeyFrame = false
-  presentationTimeUs = 250000
-sample:
-  trackType = video
-  dataHashCode = -1867110345
-  size = 2306
-  isKeyFrame = false
-  presentationTimeUs = 266666
-sample:
-  trackType = video
-  dataHashCode = 1908323737
-  size = 257
-  isKeyFrame = false
-  presentationTimeUs = 283333
-sample:
-  trackType = video
-  dataHashCode = 884063337
-  size = 2201
-  isKeyFrame = false
-  presentationTimeUs = 300000
-sample:
-  trackType = video
-  dataHashCode = -1308458590
-  size = 174
-  isKeyFrame = false
-  presentationTimeUs = 316666
-sample:
-  trackType = video
-  dataHashCode = -1686938678
-  size = 2524
-  isKeyFrame = false
-  presentationTimeUs = 333333
-sample:
-  trackType = video
-  dataHashCode = -1372845971
-  size = 171
-  isKeyFrame = false
-  presentationTimeUs = 350000
-sample:
-  trackType = video
-  dataHashCode = 1130876644
-  size = 2306
-  isKeyFrame = false
-  presentationTimeUs = 366666
-sample:
-  trackType = video
-  dataHashCode = 1707671352
-  size = 188
-  isKeyFrame = false
-  presentationTimeUs = 383333
-sample:
-  trackType = video
-  dataHashCode = 300233313
-  size = 2529
-  isKeyFrame = false
-  presentationTimeUs = 400000
-sample:
-  trackType = video
-  dataHashCode = -1284013406
-  size = 182
-  isKeyFrame = false
-  presentationTimeUs = 416666
-sample:
-  trackType = video
-  dataHashCode = -2088617828
-  size = 2047
-  isKeyFrame = false
-  presentationTimeUs = 433333
-sample:
-  trackType = video
-  dataHashCode = 2116374999
-  size = 259
-  isKeyFrame = false
-  presentationTimeUs = 450000
-sample:
-  trackType = video
-  dataHashCode = -2123019940
-  size = 2234
-  isKeyFrame = false
-  presentationTimeUs = 466666
-sample:
-  trackType = video
-  dataHashCode = 1901454757
-  size = 138
-  isKeyFrame = false
-  presentationTimeUs = 483333
-sample:
-  trackType = video
-  dataHashCode = 1576638059
-  size = 2088
-  isKeyFrame = false
-  presentationTimeUs = 500000
-sample:
-  trackType = video
-  dataHashCode = 1120133924
-  size = 151
-  isKeyFrame = false
-  presentationTimeUs = 516666
-sample:
-  trackType = video
-  dataHashCode = 264118578
-  size = 2235
-  isKeyFrame = false
-  presentationTimeUs = 533333
-sample:
-  trackType = video
-  dataHashCode = 64254117
-  size = 164
-  isKeyFrame = false
-  presentationTimeUs = 550000
-sample:
-  trackType = video
-  dataHashCode = -1000078879
-  size = 2231
-  isKeyFrame = false
-  presentationTimeUs = 566666
-sample:
-  trackType = video
-  dataHashCode = 286919946
-  size = 123
-  isKeyFrame = false
-  presentationTimeUs = 583333
-sample:
-  trackType = video
-  dataHashCode = -320312658
-  size = 2303
-  isKeyFrame = false
-  presentationTimeUs = 600000
-sample:
-  trackType = video
-  dataHashCode = 1057750590
-  size = 175
-  isKeyFrame = false
-  presentationTimeUs = 616666
-sample:
-  trackType = video
-  dataHashCode = 1961415074
-  size = 2165
-  isKeyFrame = false
-  presentationTimeUs = 633333
-sample:
-  trackType = video
-  dataHashCode = 667267023
-  size = 260
-  isKeyFrame = false
-  presentationTimeUs = 650000
-sample:
-  trackType = video
-  dataHashCode = 979033489
-  size = 1924
-  isKeyFrame = false
-  presentationTimeUs = 666666
-sample:
-  trackType = video
-  dataHashCode = -1974473017
-  size = 286
-  isKeyFrame = false
-  presentationTimeUs = 683333
-sample:
-  trackType = video
-  dataHashCode = -962519103
-  size = 1992
-  isKeyFrame = false
-  presentationTimeUs = 700000
-sample:
-  trackType = video
-  dataHashCode = -1312094075
-  size = 204
-  isKeyFrame = false
-  presentationTimeUs = 716666
-sample:
-  trackType = video
-  dataHashCode = 2068151127
-  size = 1826
-  isKeyFrame = false
-  presentationTimeUs = 733333
-sample:
-  trackType = video
-  dataHashCode = -1531967506
-  size = 284
-  isKeyFrame = false
-  presentationTimeUs = 750000
-sample:
-  trackType = video
-  dataHashCode = -778066699
-  size = 1940
-  isKeyFrame = false
-  presentationTimeUs = 766666
-sample:
-  trackType = video
-  dataHashCode = -1219952117
-  size = 129
-  isKeyFrame = false
-  presentationTimeUs = 783333
-sample:
-  trackType = video
-  dataHashCode = -1218204223
-  size = 1947
-  isKeyFrame = false
-  presentationTimeUs = 800000
-sample:
-  trackType = video
-  dataHashCode = -1816247511
-  size = 147
-  isKeyFrame = false
-  presentationTimeUs = 816666
-sample:
-  trackType = video
-  dataHashCode = 299686318
-  size = 2066
-  isKeyFrame = false
-  presentationTimeUs = 833333
-sample:
-  trackType = video
-  dataHashCode = -1520242765
-  size = 185
-  isKeyFrame = false
-  presentationTimeUs = 850000
-sample:
-  trackType = video
-  dataHashCode = -1702498409
-  size = 2159
-  isKeyFrame = false
-  presentationTimeUs = 866666
-sample:
-  trackType = video
-  dataHashCode = 345202950
-  size = 189
-  isKeyFrame = false
-  presentationTimeUs = 883333
-sample:
-  trackType = video
-  dataHashCode = 220746796
-  size = 2098
-  isKeyFrame = false
-  presentationTimeUs = 900000
-sample:
-  trackType = video
-  dataHashCode = -32341189
-  size = 159
-  isKeyFrame = false
-  presentationTimeUs = 916666
-sample:
-  trackType = video
-  dataHashCode = -1838476361
-  size = 1914
-  isKeyFrame = false
-  presentationTimeUs = 933333
-sample:
-  trackType = video
-  dataHashCode = -1322093590
-  size = 99
-  isKeyFrame = false
-  presentationTimeUs = 950000
-sample:
-  trackType = video
-  dataHashCode = -1391064751
-  size = 2168
-  isKeyFrame = false
-  presentationTimeUs = 966666
-sample:
-  trackType = video
-  dataHashCode = 1479204931
-  size = 129
-  isKeyFrame = false
-  presentationTimeUs = 983333
-sample:
-  trackType = video
-  dataHashCode = 1131230500
-  size = 2327
-  isKeyFrame = false
-  presentationTimeUs = 1000000
-sample:
-  trackType = video
-  dataHashCode = -393815961
-  size = 160
-  isKeyFrame = false
-  presentationTimeUs = 1016666
-sample:
-  trackType = video
-  dataHashCode = -242739025
-  size = 2136
-  isKeyFrame = false
-  presentationTimeUs = 1033333
-sample:
-  trackType = video
-  dataHashCode = 65238903
-  size = 163
-  isKeyFrame = false
-  presentationTimeUs = 1050000
-sample:
-  trackType = video
-  dataHashCode = 1720840922
-  size = 2043
-  isKeyFrame = false
-  presentationTimeUs = 1066666
-sample:
-  trackType = video
-  dataHashCode = -1006231050
-  size = 178
-  isKeyFrame = false
-  presentationTimeUs = 1083333
-sample:
-  trackType = video
-  dataHashCode = 1742965952
-  size = 2022
-  isKeyFrame = false
-  presentationTimeUs = 1100000
-sample:
-  trackType = video
-  dataHashCode = -971065365
-  size = 240
-  isKeyFrame = false
-  presentationTimeUs = 1116666
-sample:
-  trackType = video
-  dataHashCode = 1757434551
-  size = 1887
-  isKeyFrame = false
-  presentationTimeUs = 1133333
-sample:
-  trackType = video
-  dataHashCode = 1501849116
-  size = 252
-  isKeyFrame = false
-  presentationTimeUs = 1150000
-sample:
-  trackType = video
-  dataHashCode = 825501977
-  size = 1816
-  isKeyFrame = false
-  presentationTimeUs = 1166666
-sample:
-  trackType = video
-  dataHashCode = -1616223509
-  size = 246
-  isKeyFrame = false
-  presentationTimeUs = 1183333
-sample:
-  trackType = video
-  dataHashCode = 457119646
-  size = 1817
-  isKeyFrame = false
-  presentationTimeUs = 1200000
-sample:
-  trackType = video
-  dataHashCode = -1382929639
-  size = 146
-  isKeyFrame = false
-  presentationTimeUs = 1216666
-sample:
-  trackType = video
-  dataHashCode = -1580853131
-  size = 1929
-  isKeyFrame = false
-  presentationTimeUs = 1233333
-sample:
-  trackType = video
-  dataHashCode = 1758706551
-  size = 196
-  isKeyFrame = false
-  presentationTimeUs = 1250000
-sample:
-  trackType = video
-  dataHashCode = 207289556
-  size = 2154
-  isKeyFrame = false
-  presentationTimeUs = 1266666
-sample:
-  trackType = video
-  dataHashCode = -981284942
-  size = 182
-  isKeyFrame = false
-  presentationTimeUs = 1283333
-sample:
-  trackType = video
-  dataHashCode = 855103964
-  size = 2144
-  isKeyFrame = false
-  presentationTimeUs = 1300000
-sample:
-  trackType = video
-  dataHashCode = 380479426
-  size = 90
-  isKeyFrame = false
-  presentationTimeUs = 1316666
-sample:
-  trackType = video
-  dataHashCode = -1677996152
-  size = 2005
-  isKeyFrame = false
-  presentationTimeUs = 1333333
-sample:
-  trackType = video
-  dataHashCode = 1516852008
-  size = 156
-  isKeyFrame = false
-  presentationTimeUs = 1350000
-sample:
-  trackType = video
-  dataHashCode = -1602805193
-  size = 1772
-  isKeyFrame = false
-  presentationTimeUs = 1366666
-sample:
-  trackType = video
-  dataHashCode = -1720426556
-  size = 162
-  isKeyFrame = false
-  presentationTimeUs = 1383333
-sample:
-  trackType = video
-  dataHashCode = -1392260423
-  size = 1865
-  isKeyFrame = false
-  presentationTimeUs = 1400000
-sample:
-  trackType = video
-  dataHashCode = -1842432151
-  size = 151
-  isKeyFrame = false
-  presentationTimeUs = 1416666
-sample:
-  trackType = video
-  dataHashCode = -537063215
-  size = 1848
-  isKeyFrame = false
-  presentationTimeUs = 1433333
-sample:
-  trackType = video
-  dataHashCode = 2089388394
-  size = 206
-  isKeyFrame = false
-  presentationTimeUs = 1450000
-sample:
-  trackType = video
-  dataHashCode = -1761777019
-  size = 1934
-  isKeyFrame = false
-  presentationTimeUs = 1466666
-sample:
-  trackType = video
-  dataHashCode = 235471194
-  size = 119
-  isKeyFrame = false
-  presentationTimeUs = 1483333
-released = true

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

@@ -16,6 +16,7 @@
 package androidx.media3.transformer;
 
 import static androidx.media3.common.util.Assertions.checkNotNull;
+import static androidx.media3.common.util.Util.isRunningOnEmulator;
 import static androidx.media3.transformer.AndroidTestUtil.JPG_ASSET_URI_STRING;
 import static androidx.media3.transformer.AndroidTestUtil.MP3_ASSET_URI_STRING;
 import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_FORMAT;
@@ -26,8 +27,10 @@ import static androidx.media3.transformer.AndroidTestUtil.MP4_ASSET_WITH_INCREAS
 import static androidx.media3.transformer.AndroidTestUtil.PNG_ASSET_URI_STRING;
 import static androidx.media3.transformer.AndroidTestUtil.createOpenGlObjects;
 import static androidx.media3.transformer.AndroidTestUtil.generateTextureFromBitmap;
+import static androidx.media3.transformer.AndroidTestUtil.recordTestSkipped;
 import static com.google.common.truth.Truth.assertThat;
 import static org.junit.Assert.assertThrows;
+import static org.junit.Assume.assumeTrue;
 
 import android.content.Context;
 import android.graphics.Bitmap;
@@ -49,6 +52,7 @@ import androidx.media3.common.audio.ChannelMixingAudioProcessor;
 import androidx.media3.common.audio.ChannelMixingMatrix;
 import androidx.media3.common.audio.SonicAudioProcessor;
 import androidx.media3.common.util.GlUtil;
+import androidx.media3.common.util.Util;
 import androidx.media3.datasource.DataSourceBitmapLoader;
 import androidx.media3.effect.Contrast;
 import androidx.media3.effect.DefaultGlObjectsProvider;
@@ -437,6 +441,43 @@ public class TransformerEndToEndTest {
     assertThat(result.exportResult.durationMs).isAtMost(clippingEndMs - clippingStartMs);
   }
 
+  @Test
+  public void clippedMedia_trimOptimizationEnabled_completesWithOptimizationApplied()
+      throws Exception {
+    String testId = "clippedMedia_trimOptimizationEnabled_completesWithOptimizationApplied";
+    if (!isRunningOnEmulator()) {
+      // The trim optimization is only guaranteed to work on emulator for this file.
+      recordTestSkipped(context, testId, /* reason= */ "Emulator only test");
+      assumeTrue(false);
+    }
+    if (Util.SDK_INT == 26) {
+      // MediaCodec returns a segmentation fault fails at this SDK level on emulators.
+      recordTestSkipped(context, testId, /* reason= */ "SDK 26 not supported.");
+      assumeTrue(false);
+    }
+    Transformer transformer =
+        new Transformer.Builder(context).experimentalSetTrimOptimizationEnabled(true).build();
+    MediaItem mediaItem =
+        new MediaItem.Builder()
+            .setUri("asset:///media/mp4/crow_emulator_transformer_output.mp4")
+            .setClippingConfiguration(
+                new MediaItem.ClippingConfiguration.Builder()
+                    .setStartPositionMs(500)
+                    .setEndPositionMs(2500)
+                    .build())
+            .build();
+    EditedMediaItem editedMediaItem = new EditedMediaItem.Builder(mediaItem).build();
+
+    ExportTestResult result =
+        new TransformerAndroidTestRunner.Builder(context, transformer)
+            .build()
+            .run(testId, editedMediaItem);
+
+    assertThat(result.exportResult.optimizationResult)
+        .isEqualTo(ExportResult.OPTIMIZATION_SUCCEEDED);
+    assertThat(result.exportResult.durationMs).isAtMost(2000);
+  }
+
   @Test
   public void videoEncoderFormatUnsupported_completesWithError() throws Exception {
     String testId = "videoEncoderFormatUnsupported_completesWithError";

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

@@ -256,6 +256,7 @@ public final class ExportResult {
       width = C.LENGTH_UNSET;
       videoFrameCount = 0;
       videoEncoderName = null;
+      optimizationResult = OPTIMIZATION_NONE;
       exportException = null;
     }
   }
@@ -418,6 +419,7 @@ public final class ExportResult {
         .setWidth(width)
         .setVideoFrameCount(videoFrameCount)
         .setVideoEncoderName(videoEncoderName)
+        .setOptimizationResult(optimizationResult)
         .setExportException(exportException);
   }
 
@@ -443,6 +445,7 @@ public final class ExportResult {
         && width == result.width
         && videoFrameCount == result.videoFrameCount
         && Objects.equals(videoEncoderName, result.videoEncoderName)
+        && optimizationResult == result.optimizationResult
         && Objects.equals(exportException, result.exportException);
   }
 
@@ -461,6 +464,7 @@ public final class ExportResult {
     result = 31 * result + width;
     result = 31 * result + videoFrameCount;
     result = 31 * result + Objects.hashCode(videoEncoderName);
+    result = 31 * result + optimizationResult;
     result = 31 * result + Objects.hashCode(exportException);
     return result;
   }

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

@@ -19,7 +19,9 @@ package androidx.media3.transformer;
 import static androidx.media3.common.util.Assertions.checkArgument;
 import static androidx.media3.common.util.Assertions.checkNotNull;
 import static androidx.media3.common.util.Assertions.checkState;
+import static androidx.media3.common.util.Util.areEqual;
 import static androidx.media3.common.util.Util.contains;
+import static androidx.media3.common.util.Util.usToMs;
 import static java.lang.Math.max;
 import static java.lang.annotation.ElementType.TYPE_USE;
 import static java.util.concurrent.TimeUnit.MILLISECONDS;
@@ -57,22 +59,21 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
   @Documented
   @Retention(RetentionPolicy.SOURCE)
   @Target(TYPE_USE)
-  @IntDef({MUXER_MODE_DEFAULT, MUXER_MODE_MUX_PARTIAL_VIDEO, MUXER_MODE_APPEND_VIDEO})
+  @IntDef({MUXER_MODE_DEFAULT, MUXER_MODE_MUX_PARTIAL, MUXER_MODE_APPEND})
   public @interface MuxerMode {}
 
   /** The default muxer mode. */
   public static final int MUXER_MODE_DEFAULT = 0;
 
   /**
-   * Used for muxing a partial video. The video {@link TrackInfo} is kept the same when {@linkplain
+   * Used for muxing a partial track(s). The {@link TrackInfo} is kept the same when {@linkplain
-   * #changeToAppendVideoMode() transitioning} to {@link #MUXER_MODE_APPEND_VIDEO} after finishing
+   * #changeToAppendMode() transitioning} to {@link #MUXER_MODE_APPEND} after finishing muxing
-   * muxing partial video. Only one video track can be {@link #addTrackFormat(Format) added} in this
+   * partial tracks.
-   * mode.
    */
-  public static final int MUXER_MODE_MUX_PARTIAL_VIDEO = 1;
+  public static final int MUXER_MODE_MUX_PARTIAL = 1;
 
-  /** Used for appending the remaining video samples with the previously muxed partial video. */
+  /** Used for appending the remaining samples with the previously muxed partial file. */
-  public static final int MUXER_MODE_APPEND_VIDEO = 2;
+  public static final int MUXER_MODE_APPEND = 2;
 
   private static final String TIMER_THREAD_NAME = "Muxer:Timer";
   private static final String MUXER_TIMEOUT_ERROR_FORMAT_STRING =
@@ -111,7 +112,11 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
   private @MonotonicNonNull Muxer muxer;
   private @MuxerMode int muxerMode;
 
+  // Read by any thread, only written to on the transformerInternal thread.
   private volatile boolean muxedPartialVideo;
+  // Read by any thread, only written to on the transformerInternal thread.
+  private volatile boolean muxedPartialAudio;
+
   private volatile int additionalRotationDegrees;
   private volatile int trackCount;
 
@@ -122,14 +127,14 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
    * @param muxerFactory A {@link Muxer.Factory} to create a {@link Muxer}.
    * @param listener A {@link MuxerWrapper.Listener}.
    * @param muxerMode The {@link MuxerMode}. The initial mode must be {@link #MUXER_MODE_DEFAULT} or
-   *     {@link #MUXER_MODE_MUX_PARTIAL_VIDEO}.
+   *     {@link #MUXER_MODE_MUX_PARTIAL}.
    */
   public MuxerWrapper(
       String outputPath, Muxer.Factory muxerFactory, Listener listener, @MuxerMode int muxerMode) {
     this.outputPath = outputPath;
     this.muxerFactory = muxerFactory;
     this.listener = listener;
-    checkArgument(muxerMode == MUXER_MODE_DEFAULT || muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO);
+    checkArgument(muxerMode == MUXER_MODE_DEFAULT || muxerMode == MUXER_MODE_MUX_PARTIAL);
     this.muxerMode = muxerMode;
     trackTypeToInfo = new SparseArray<>();
     previousTrackType = C.TRACK_TYPE_NONE;
@@ -137,15 +142,15 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
   }
 
   /**
-   * Changes {@link MuxerMode} to {@link #MUXER_MODE_APPEND_VIDEO}.
+   * Changes {@link MuxerMode} to {@link #MUXER_MODE_APPEND}.
    *
-   * <p>This method must be called only after partial video is muxed using {@link
+   * <p>This method must be called only after partial file is muxed using {@link
-   * #MUXER_MODE_MUX_PARTIAL_VIDEO}.
+   * #MUXER_MODE_MUX_PARTIAL}.
    */
-  public void changeToAppendVideoMode() {
+  public void changeToAppendMode() {
-    checkState(muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO);
+    checkState(muxerMode == MUXER_MODE_MUX_PARTIAL);
 
-    muxerMode = MUXER_MODE_APPEND_VIDEO;
+    muxerMode = MUXER_MODE_APPEND;
   }
 
   /**
@@ -181,15 +186,9 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
    *     before calling this method.
    */
   public void setTrackCount(@IntRange(from = 1) int trackCount) {
-    if (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO || muxerMode == MUXER_MODE_APPEND_VIDEO) {
+    if (muxerMode == MUXER_MODE_APPEND) {
-      checkArgument(
-          trackCount == 1,
-          "Only one video track can be added in MUXER_MODE_MUX_PARTIAL_VIDEO and"
-              + " MUXER_MODE_APPEND_VIDEO");
-      if (muxerMode == MUXER_MODE_APPEND_VIDEO) {
       return;
     }
-    }
     checkState(
         trackTypeToInfo.size() == 0,
         "The track count cannot be changed after adding track formats.");
@@ -219,11 +218,11 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
    *
    * <p>{@link Muxer#addMetadata(Metadata)} is called if the {@link Format#metadata} is present.
    *
-   * @param format The {@link Format} to be added. In {@link #MUXER_MODE_APPEND_VIDEO} mode, the
+   * @param format The {@link Format} to be added. In {@link #MUXER_MODE_APPEND} mode, the added
-   *     added {@link Format} must match the existing {@link Format} set when the muxer was in
+   *     {@link Format} must match the existing {@link Format} set when the muxer was in {@link
-   *     {@link #MUXER_MODE_MUX_PARTIAL_VIDEO} mode.
+   *     #MUXER_MODE_MUX_PARTIAL} mode.
    * @throws IllegalArgumentException If the format is unsupported or if it does not match the
-   *     existing format in {@link #MUXER_MODE_APPEND_VIDEO} mode.
+   *     existing format in {@link #MUXER_MODE_APPEND} mode.
    * @throws IllegalStateException If the number of formats added exceeds the {@linkplain
    *     #setTrackCount track count}, if {@link #setTrackCount(int)} has not been called or if there
    *     is already a track of that {@link C.TrackType}.
@@ -231,36 +230,43 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
    *     the track.
    */
   public void addTrackFormat(Format format) throws Muxer.MuxerException {
-    if (muxerMode == MUXER_MODE_APPEND_VIDEO) {
-      checkState(contains(trackTypeToInfo, C.TRACK_TYPE_VIDEO));
-      TrackInfo videoTrackInfo = trackTypeToInfo.get(C.TRACK_TYPE_VIDEO);
-
-      // Ensure that video formats are the same. Some fields like codecs, averageBitrate, framerate,
-      // etc, don't match exactly in the Extractor output format and the Encoder output
-      // format but these fields can be ignored.
-      // TODO: b/308180225 - Compare Format.colorInfo as well.
-      Format existingFormat = videoTrackInfo.format;
-      checkArgument(Util.areEqual(existingFormat.sampleMimeType, format.sampleMimeType));
-      checkArgument(existingFormat.width == format.width);
-      checkArgument(existingFormat.height == format.height);
-      checkArgument(existingFormat.initializationDataEquals(format));
-
-      checkNotNull(muxer);
-      resetAbortTimer();
-      return;
-    } else if (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO) {
-      checkArgument(MimeTypes.isVideo(format.sampleMimeType));
-    }
-
-    int trackCount = this.trackCount;
-    checkState(trackCount > 0, "The track count should be set before the formats are added.");
-    checkState(trackTypeToInfo.size() < trackCount, "All track formats have already been added.");
     @Nullable String sampleMimeType = format.sampleMimeType;
     @C.TrackType int trackType = MimeTypes.getTrackType(sampleMimeType);
     checkArgument(
         trackType == C.TRACK_TYPE_AUDIO || trackType == C.TRACK_TYPE_VIDEO,
         "Unsupported track format: " + sampleMimeType);
+    if (muxerMode == MUXER_MODE_APPEND) {
+      if (trackType == C.TRACK_TYPE_VIDEO) {
+        checkState(contains(trackTypeToInfo, C.TRACK_TYPE_VIDEO));
+        TrackInfo videoTrackInfo = trackTypeToInfo.get(C.TRACK_TYPE_VIDEO);
 
+        // Ensure that video formats are the same. Some fields like codecs, averageBitrate,
+        // framerate, etc, don't match exactly in the Extractor output format and the Encoder output
+        // format but these fields can be ignored.
+        // TODO: b/308180225 - Compare Format.colorInfo as well.
+        Format existingFormat = videoTrackInfo.format;
+        checkArgument(areEqual(existingFormat.sampleMimeType, format.sampleMimeType));
+        checkArgument(existingFormat.width == format.width);
+        checkArgument(existingFormat.height == format.height);
+        checkArgument(existingFormat.initializationDataEquals(format));
+      } else if (trackType == C.TRACK_TYPE_AUDIO) {
+        checkState(contains(trackTypeToInfo, C.TRACK_TYPE_AUDIO));
+        TrackInfo audioTrackInfo = trackTypeToInfo.get(C.TRACK_TYPE_AUDIO);
+
+        Format existingFormat = audioTrackInfo.format;
+        checkArgument(areEqual(existingFormat.sampleMimeType, format.sampleMimeType));
+        checkArgument(existingFormat.channelCount == format.channelCount);
+        checkArgument(existingFormat.sampleRate == format.sampleRate);
+        checkArgument(existingFormat.initializationDataEquals(format));
+      }
+      checkNotNull(muxer);
+      resetAbortTimer();
+      return;
+    }
+
+    int trackCount = this.trackCount;
+    checkState(trackCount > 0, "The track count should be set before the formats are added.");
+    checkState(trackTypeToInfo.size() < trackCount, "All track formats have already been added.");
     checkState(
         !contains(trackTypeToInfo, trackType), "There is already a track of type " + trackType);
 
@@ -362,8 +368,12 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
       DebugTraceUtil.logEvent(DebugTraceUtil.EVENT_MUXER_TRACK_ENDED_AUDIO, trackInfo.timeUs);
     }
 
-    if (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO) {
+    if (muxerMode == MUXER_MODE_MUX_PARTIAL) {
+      if (trackType == C.TRACK_TYPE_VIDEO) {
         muxedPartialVideo = true;
+      } else if (trackType == C.TRACK_TYPE_AUDIO) {
+        muxedPartialAudio = true;
+      }
     } else {
       trackTypeToInfo.delete(trackType);
       if (trackTypeToInfo.size() == 0) {
@@ -371,8 +381,10 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
       }
     }
 
-    if (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO && muxedPartialVideo) {
+    if (muxerMode == MUXER_MODE_MUX_PARTIAL
-      listener.onEnded(Util.usToMs(maxEndedTrackTimeUs), getCurrentOutputSizeBytes());
+        && muxedPartialVideo
+        && (muxedPartialAudio || trackCount == 1)) {
+      listener.onEnded(usToMs(maxEndedTrackTimeUs), getCurrentOutputSizeBytes());
       if (abortScheduledFuture != null) {
         abortScheduledFuture.cancel(/* mayInterruptIfRunning= */ false);
       }
@@ -380,26 +392,29 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
     }
 
     if (isEnded) {
-      listener.onEnded(Util.usToMs(maxEndedTrackTimeUs), getCurrentOutputSizeBytes());
+      listener.onEnded(usToMs(maxEndedTrackTimeUs), getCurrentOutputSizeBytes());
       abortScheduledExecutorService.shutdownNow();
     }
   }
 
   /**
-   * Returns whether all the tracks are {@linkplain #endTrack(int) ended} or a partial video is
+   * Returns whether all the tracks are {@linkplain #endTrack(int) ended} or a partial file is
-   * completely muxed using {@link #MUXER_MODE_MUX_PARTIAL_VIDEO}.
+   * completely muxed using {@link #MUXER_MODE_MUX_PARTIAL}.
    */
   public boolean isEnded() {
-    return isEnded || (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO && muxedPartialVideo);
+    return isEnded
+        || (muxerMode == MUXER_MODE_MUX_PARTIAL
+            && muxedPartialVideo
+            && (muxedPartialAudio || trackCount == 1));
   }
 
   /**
    * Finishes writing the output and releases any resources associated with muxing.
    *
-   * <p>When this method is called in {@link #MUXER_MODE_MUX_PARTIAL_VIDEO} mode, the resources are
+   * <p>When this method is called in {@link #MUXER_MODE_MUX_PARTIAL} mode, the resources are not
-   * not released and the {@link MuxerWrapper} can be reused after {@link #changeToAppendVideoMode()
+   * released and the {@link MuxerWrapper} can be reused after {@link #changeToAppendMode() changing
-   * changing mode} to {@link #MUXER_MODE_APPEND_VIDEO} mode. In all other modes the {@link
+   * mode} to {@link #MUXER_MODE_APPEND}. In all other modes the {@link MuxerWrapper} cannot be used
-   * MuxerWrapper} cannot be used anymore once this method has been called.
+   * anymore once this method has been called.
    *
    * @param forCancellation Whether the reason for releasing the resources is the transformation
    *     cancellation.
@@ -407,7 +422,7 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
    *     and {@code forCancellation} is false.
    */
   public void release(boolean forCancellation) throws Muxer.MuxerException {
-    if (muxerMode == MUXER_MODE_MUX_PARTIAL_VIDEO && !forCancellation) {
+    if (muxerMode == MUXER_MODE_MUX_PARTIAL && !forCancellation) {
       return;
     }
     isReady = false;

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

@@ -287,7 +287,7 @@ public final class Transformer {
       return this;
     }
 
-    // TODO: b/304476154 - Support audio and progress updates in trim optimization.
+    // TODO: b/304476154 - Support progress updates in trim optimization.
     /**
      * Sets whether to attempt to optimize trims from the start of the {@link EditedMediaItem} by
      * transcoding as little of the file as possible and transmuxing the rest.
@@ -298,7 +298,6 @@ public final class Transformer {
      *   <li>Only supported for single-asset (i.e. only one {@link EditedMediaItem} in the whole
      *       {@link Composition}) exports of mp4 files.
      *   <li>Not guaranteed to work with any effects.
-     *   <li>Video track only (removes audio from the file).
      *   <li>Progress updates will be unavailable.
      * </ul>
      *
@@ -708,8 +707,8 @@ public final class Transformer {
     TRANSFORMER_STATE_PROCESS_REMAINING_VIDEO,
     TRANSFORMER_STATE_PROCESS_AUDIO,
     TRANSFORMER_STATE_COPY_OUTPUT,
-    TRANSFORMER_STATE_TRANSCODE_VIDEO_START,
+    TRANSFORMER_STATE_PROCESS_MEDIA_START,
-    TRANSFORMER_STATE_TRANSMUX_REMAINING_VIDEO
+    TRANSFORMER_STATE_REMUX_REMAINING_MEDIA
   })
   private @interface TransformerState {}
 
@@ -718,9 +717,11 @@ public final class Transformer {
   private static final int TRANSFORMER_STATE_PROCESS_REMAINING_VIDEO = 2;
   private static final int TRANSFORMER_STATE_PROCESS_AUDIO = 3;
   private static final int TRANSFORMER_STATE_COPY_OUTPUT = 4;
-  private static final int TRANSFORMER_STATE_TRANSCODE_VIDEO_START = 5;
+  private static final int TRANSFORMER_STATE_PROCESS_MEDIA_START = 5;
-  private static final int TRANSFORMER_STATE_TRANSMUX_REMAINING_VIDEO = 6;
+  private static final int TRANSFORMER_STATE_REMUX_REMAINING_MEDIA = 6;
 
+  // TODO: b/304476154 - Calculate duration based on sample rate from audio format.
+  private static final int MAX_ENCODED_AUDIO_BUFFER_DURATION_US = 25_000;
   private final Context context;
   private final TransformationRequest transformationRequest;
   private final ImmutableList<AudioProcessor> audioProcessors;
@@ -925,7 +926,7 @@ public final class Transformer {
           componentListener,
           /* initialTimestampOffsetUs= */ 0);
     } else {
-      transcodeVideoBeforeFirstSyncSampleAfterTrimStartTime();
+      processMediaBeforeFirstSyncSampleAfterTrimStartTime();
     }
   }
 
@@ -1139,7 +1140,7 @@ public final class Transformer {
                     checkNotNull(outputFilePath),
                     muxerFactory,
                     componentListener,
-                    MuxerWrapper.MUXER_MODE_MUX_PARTIAL_VIDEO);
+                    MuxerWrapper.MUXER_MODE_MUX_PARTIAL);
 
             startInternal(
                 TransmuxTranscodeHelper.createVideoOnlyComposition(
@@ -1169,7 +1170,7 @@ public final class Transformer {
             resumeMetadata);
 
     checkNotNull(remuxingMuxerWrapper);
-    remuxingMuxerWrapper.changeToAppendVideoMode();
+    remuxingMuxerWrapper.changeToAppendMode();
 
     startInternal(
         videoOnlyComposition,
@@ -1218,11 +1219,12 @@ public final class Transformer {
         applicationHandler::post);
   }
 
-  private void transcodeVideoBeforeFirstSyncSampleAfterTrimStartTime() {
+  private void processMediaBeforeFirstSyncSampleAfterTrimStartTime() {
-    transformerState = TRANSFORMER_STATE_TRANSCODE_VIDEO_START;
+    transformerState = TRANSFORMER_STATE_PROCESS_MEDIA_START;
     MediaItem firstMediaItem =
         checkNotNull(composition).sequences.get(0).editedMediaItems.get(0).mediaItem;
     long trimStartTimeUs = firstMediaItem.clippingConfiguration.startPositionUs;
+    long trimEndTimeUs = firstMediaItem.clippingConfiguration.endPositionUs;
     getMp4MetadataInfoFuture =
         TransmuxTranscodeHelper.getMp4MetadataInfo(
             context,
@@ -1233,16 +1235,21 @@ public final class Transformer {
         new FutureCallback<Mp4MetadataInfo>() {
           @Override
           public void onSuccess(Mp4MetadataInfo mp4MetadataInfo) {
-            if (mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs == C.TIME_UNSET) {
+            if (mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs == C.TIME_UNSET
+                || (trimEndTimeUs != C.TIME_END_OF_SOURCE
+                    && trimEndTimeUs < mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs)) {
               exportResultBuilder.setOptimizationResult(OPTIMIZATION_FAILED_NO_VIDEO_TRACK_TO_TRIM);
               processFullInput();
               return;
             }
-            if (mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs == trimStartTimeUs) {
+            // Ensure there is an audio sample to mux between the two clip times to prevent
+            // Transformer from hanging because it received an audio track but no audio samples.
+            if (mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs - trimStartTimeUs
+                < MAX_ENCODED_AUDIO_BUFFER_DURATION_US) {
               Transformer.this.composition =
                   buildNewCompositionWithClipTimes(
                       composition,
-                      trimStartTimeUs,
+                      mp4MetadataInfo.firstSyncSampleTimestampUsAfterTimeUs,
                       firstMediaItem.clippingConfiguration.endPositionUs,
                       mp4MetadataInfo.durationUs,
                       /* startsAtKeyFrame= */ true);
@@ -1266,7 +1273,7 @@ public final class Transformer {
                     checkNotNull(outputFilePath),
                     muxerFactory,
                     componentListener,
-                    MuxerWrapper.MUXER_MODE_MUX_PARTIAL_VIDEO);
+                    MuxerWrapper.MUXER_MODE_MUX_PARTIAL);
             startInternal(
                 trancodeComposition,
                 remuxingMuxerWrapper,
@@ -1283,8 +1290,8 @@ public final class Transformer {
         applicationHandler::post);
   }
 
-  private void transmuxRemainingVideo() {
+  private void remuxRemainingMedia() {
-    transformerState = TRANSFORMER_STATE_TRANSMUX_REMAINING_VIDEO;
+    transformerState = TRANSFORMER_STATE_REMUX_REMAINING_MEDIA;
     // TODO: b/304476154 - check original file format against transcode file format here to fail
     //  fast if necessary.
     MediaItem firstMediaItem =
@@ -1300,7 +1307,7 @@ public final class Transformer {
             mp4MetadataInfo.durationUs,
             /* startsAtKeyFrame= */ true);
     checkNotNull(remuxingMuxerWrapper);
-    remuxingMuxerWrapper.changeToAppendVideoMode();
+    remuxingMuxerWrapper.changeToAppendMode();
     startInternal(
         transmuxComposition,
         remuxingMuxerWrapper,
@@ -1410,9 +1417,9 @@ public final class Transformer {
         processAudio();
       } else if (transformerState == TRANSFORMER_STATE_PROCESS_AUDIO) {
         copyOutput();
-      } else if (transformerState == TRANSFORMER_STATE_TRANSCODE_VIDEO_START) {
+      } else if (transformerState == TRANSFORMER_STATE_PROCESS_MEDIA_START) {
-        transmuxRemainingVideo();
+        remuxRemainingMedia();
-      } else if (transformerState == TRANSFORMER_STATE_TRANSMUX_REMAINING_VIDEO) {
+      } else if (transformerState == TRANSFORMER_STATE_REMUX_REMAINING_MEDIA) {
         exportResultBuilder.setOptimizationResult(ExportResult.OPTIMIZATION_SUCCEEDED);
         onExportCompletedWithSuccess();
       } else {

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

@@ -101,8 +101,6 @@ import java.util.List;
             .buildUpon()
             .setMediaItem(mediaItem)
             .setDurationUs(mediaDurationUs)
-            // TODO: b/304476154 - Support audio in trim optimization.
-            .setRemoveAudio(true)
             .build();
 
     return oldComposition

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

@@ -196,14 +196,12 @@ public final class MediaItemExportTest {
     ExportResult exportResult = TransformerTestRunner.runLooper(transformer);
 
     assertThat(exportResult.optimizationResult).isEqualTo(ExportResult.OPTIMIZATION_NONE);
-    // TODO: b/304476154 - When trim optimization supports audio, remove trim optimization specific
-    //   file and use the pre-existing clipped file made from normal export path.
     DumpFileAsserts.assertOutput(
         context,
         muxerFactory.getCreatedMuxer(),
         getDumpFileName(
             /* originalFileName= */ FILE_AUDIO_VIDEO_INCREASING_TIMESTAMPS_15S,
-            /* modifications...= */ "trimOptimizedClippedAtKeyFrame"));
+            /* modifications...= */ "clipped"));
   }
 
   @Test

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

@@ -17,7 +17,7 @@ package androidx.media3.transformer;
 
 import static androidx.media3.common.MimeTypes.AUDIO_AAC;
 import static androidx.media3.common.MimeTypes.VIDEO_H264;
-import static androidx.media3.transformer.MuxerWrapper.MUXER_MODE_MUX_PARTIAL_VIDEO;
+import static androidx.media3.transformer.MuxerWrapper.MUXER_MODE_MUX_PARTIAL;
 import static com.google.common.truth.Truth.assertThat;
 import static org.junit.Assert.assertThrows;
 
@@ -66,7 +66,7 @@ public class MuxerWrapperTest {
   }
 
   @Test
-  public void changeToAppendVideoMode_afterDefaultMode_throws() throws Exception {
+  public void changeToAppendMode_afterDefaultMode_throws() throws Exception {
     muxerWrapper =
         new MuxerWrapper(
             temporaryFolder.newFile().getPath(),
@@ -74,86 +74,62 @@ public class MuxerWrapperTest {
             new NoOpMuxerListenerImpl(),
             MuxerWrapper.MUXER_MODE_DEFAULT);
 
-    assertThrows(IllegalStateException.class, muxerWrapper::changeToAppendVideoMode);
+    assertThrows(IllegalStateException.class, muxerWrapper::changeToAppendMode);
   }
 
   @Test
-  public void setTrackCount_toTwoInMuxPartialVideoMode_throws() throws Exception {
+  public void addTrackFormat_withSameVideoFormatInAppendMode_doesNotThrow() throws Exception {
     muxerWrapper =
         new MuxerWrapper(
             temporaryFolder.newFile().getPath(),
             new DefaultMuxer.Factory(),
             new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
+            MUXER_MODE_MUX_PARTIAL);
 
-    assertThrows(IllegalArgumentException.class, () -> muxerWrapper.setTrackCount(2));
-  }
-
-  @Test
-  public void setTrackCount_toTwoInAppendVideoMode_throws() throws Exception {
-    muxerWrapper =
-        new MuxerWrapper(
-            temporaryFolder.newFile().getPath(),
-            new DefaultMuxer.Factory(),
-            new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
     muxerWrapper.setTrackCount(1);
     muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
     muxerWrapper.writeSample(
         C.TRACK_TYPE_VIDEO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
     muxerWrapper.endTrack(C.TRACK_TYPE_VIDEO);
-    muxerWrapper.changeToAppendVideoMode();
+    muxerWrapper.changeToAppendMode();
-
-    assertThrows(IllegalArgumentException.class, () -> muxerWrapper.setTrackCount(2));
-  }
-
-  @Test
-  public void addTrackFormat_withAudioFormatInMuxPartialVideoMode_throws() throws Exception {
-    muxerWrapper =
-        new MuxerWrapper(
-            temporaryFolder.newFile().getPath(),
-            new DefaultMuxer.Factory(),
-            new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
     muxerWrapper.setTrackCount(1);
 
-    assertThrows(
+    muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
-        IllegalArgumentException.class, () -> muxerWrapper.addTrackFormat(FAKE_AUDIO_TRACK_FORMAT));
   }
 
   @Test
-  public void addTrackFormat_withSameVideoFormatInAppendVideoMode_doesNotThrow() throws Exception {
+  public void addTrackFormat_withSameAudioFormatInAppendMode_doesNotThrow() throws Exception {
     muxerWrapper =
         new MuxerWrapper(
             temporaryFolder.newFile().getPath(),
             new DefaultMuxer.Factory(),
             new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
+            MUXER_MODE_MUX_PARTIAL);
+    muxerWrapper.setTrackCount(1);
+    muxerWrapper.addTrackFormat(FAKE_AUDIO_TRACK_FORMAT);
+    muxerWrapper.writeSample(
+        C.TRACK_TYPE_AUDIO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
+    muxerWrapper.endTrack(C.TRACK_TYPE_AUDIO);
+    muxerWrapper.changeToAppendMode();
+    muxerWrapper.setTrackCount(1);
+
+    muxerWrapper.addTrackFormat(FAKE_AUDIO_TRACK_FORMAT);
+  }
+
+  @Test
+  public void addTrackFormat_withDifferentVideoFormatInAppendMode_throws() throws Exception {
+    muxerWrapper =
+        new MuxerWrapper(
+            temporaryFolder.newFile().getPath(),
+            new DefaultMuxer.Factory(),
+            new NoOpMuxerListenerImpl(),
+            MUXER_MODE_MUX_PARTIAL);
     muxerWrapper.setTrackCount(1);
     muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
     muxerWrapper.writeSample(
         C.TRACK_TYPE_VIDEO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
     muxerWrapper.endTrack(C.TRACK_TYPE_VIDEO);
-    muxerWrapper.changeToAppendVideoMode();
+    muxerWrapper.changeToAppendMode();
-    muxerWrapper.setTrackCount(1);
-
-    muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
-  }
-
-  @Test
-  public void addTrackFormat_withDifferentVideoFormatInAppendVideoMode_throws() throws Exception {
-    muxerWrapper =
-        new MuxerWrapper(
-            temporaryFolder.newFile().getPath(),
-            new DefaultMuxer.Factory(),
-            new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
-    muxerWrapper.setTrackCount(1);
-    muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
-    muxerWrapper.writeSample(
-        C.TRACK_TYPE_VIDEO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
-    muxerWrapper.endTrack(C.TRACK_TYPE_VIDEO);
-    muxerWrapper.changeToAppendVideoMode();
     muxerWrapper.setTrackCount(1);
     Format differentVideoFormat = FAKE_VIDEO_TRACK_FORMAT.buildUpon().setHeight(5000).build();
 
@@ -161,6 +137,27 @@ public class MuxerWrapperTest {
         IllegalArgumentException.class, () -> muxerWrapper.addTrackFormat(differentVideoFormat));
   }
 
+  @Test
+  public void addTrackFormat_withDifferentAudioFormatInAppendMode_throws() throws Exception {
+    muxerWrapper =
+        new MuxerWrapper(
+            temporaryFolder.newFile().getPath(),
+            new DefaultMuxer.Factory(),
+            new NoOpMuxerListenerImpl(),
+            MUXER_MODE_MUX_PARTIAL);
+    muxerWrapper.setTrackCount(1);
+    muxerWrapper.addTrackFormat(FAKE_AUDIO_TRACK_FORMAT);
+    muxerWrapper.writeSample(
+        C.TRACK_TYPE_AUDIO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
+    muxerWrapper.endTrack(C.TRACK_TYPE_AUDIO);
+    muxerWrapper.changeToAppendMode();
+    muxerWrapper.setTrackCount(1);
+    Format differentAudioFormat = FAKE_AUDIO_TRACK_FORMAT.buildUpon().setSampleRate(48000).build();
+
+    assertThrows(
+        IllegalArgumentException.class, () -> muxerWrapper.addTrackFormat(differentAudioFormat));
+  }
+
   @Test
   public void isEnded_afterPartialVideoMuxed_returnsTrue() throws Exception {
     muxerWrapper =
@@ -168,7 +165,7 @@ public class MuxerWrapperTest {
             temporaryFolder.newFile().getPath(),
             new DefaultMuxer.Factory(),
             new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
+            MUXER_MODE_MUX_PARTIAL);
     muxerWrapper.setTrackCount(1);
     muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
     muxerWrapper.writeSample(
@@ -179,19 +176,44 @@ public class MuxerWrapperTest {
   }
 
   @Test
-  public void isEnded_afterStartingAppendVideo_returnsFalse() throws Exception {
+  public void isEnded_afterPartialAudioAndVideoMuxed_returnsTrue() throws Exception {
     muxerWrapper =
         new MuxerWrapper(
             temporaryFolder.newFile().getPath(),
             new DefaultMuxer.Factory(),
             new NoOpMuxerListenerImpl(),
-            MUXER_MODE_MUX_PARTIAL_VIDEO);
+            MUXER_MODE_MUX_PARTIAL);
+
+    muxerWrapper.setTrackCount(2);
+    muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
+    muxerWrapper.addTrackFormat(FAKE_AUDIO_TRACK_FORMAT);
+    muxerWrapper.writeSample(
+        C.TRACK_TYPE_VIDEO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
+    muxerWrapper.endTrack(C.TRACK_TYPE_VIDEO);
+
+    assertThat(muxerWrapper.isEnded()).isFalse();
+
+    muxerWrapper.writeSample(
+        C.TRACK_TYPE_AUDIO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
+    muxerWrapper.endTrack(C.TRACK_TYPE_AUDIO);
+
+    assertThat(muxerWrapper.isEnded()).isTrue();
+  }
+
+  @Test
+  public void isEnded_afterEnteringAppendMode_returnsFalse() throws Exception {
+    muxerWrapper =
+        new MuxerWrapper(
+            temporaryFolder.newFile().getPath(),
+            new DefaultMuxer.Factory(),
+            new NoOpMuxerListenerImpl(),
+            MUXER_MODE_MUX_PARTIAL);
     muxerWrapper.setTrackCount(1);
     muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);
     muxerWrapper.writeSample(
         C.TRACK_TYPE_VIDEO, FAKE_SAMPLE, /* isKeyFrame= */ true, /* presentationTimeUs= */ 0);
     muxerWrapper.endTrack(C.TRACK_TYPE_VIDEO);
-    muxerWrapper.changeToAppendVideoMode();
+    muxerWrapper.changeToAppendMode();
     muxerWrapper.setTrackCount(1);
     muxerWrapper.addTrackFormat(FAKE_VIDEO_TRACK_FORMAT);