Allow setting VideoFrameMetadataListener on CompositionPlayer

PiperOrigin-RevId: 743607386

RELEASENOTES.md

@@ -36,6 +36,9 @@
 *   DataSource:
 *   Audio:
     *   Allow constant power upmixing/downmixing in DefaultAudioMixer.
+    *   Fix offload issue where position might get stuck when playing a playlist
+        of short content
+        ([#1920](https://github.com/androidx/media/issues/1920)).
 *   Video:
     *   Add experimental `ExoPlayer` API to include the
         `MediaCodec.BUFFER_FLAG_DECODE_ONLY` flag when queuing decode-only input

libraries/datasource/src/androidTest/java/androidx/media3/datasource/HttpEngineDataSourceContractTest.java

@@ -63,7 +63,7 @@ public class HttpEngineDataSourceContractTest extends DataSourceContractTest {
         new HttpEngine.Builder(ApplicationProvider.getApplicationContext()).build();
     return new HttpEngineDataSource.Factory(httpEngine, executorService)
         // Ensure that 'resource not found' tests fail fast (b/403179253).
-        .setConnectionTimeoutMs(400)
+        .setConnectionTimeoutMs(600)
         .setReadTimeoutMs(400)
         .createDataSource();
   }

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/ExoPlayerImplInternal.java

@@ -2820,6 +2820,10 @@ import java.util.Objects;
   private void maybeUpdateOffloadScheduling() {
     // If playing period is audio-only with offload mode preference to enable, then offload
     // scheduling should be enabled.
+    if (queue.getPlayingPeriod() != queue.getReadingPeriod()) {
+      // Do not enable offload scheduling when starting to process the next media item.
+      return;
+    }
     @Nullable MediaPeriodHolder playingPeriodHolder = queue.getPlayingPeriod();
     if (playingPeriodHolder != null) {
       TrackSelectorResult trackSelectorResult = playingPeriodHolder.getTrackSelectorResult();

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/audio/DefaultAudioSink.java

@@ -102,6 +102,11 @@ public final class DefaultAudioSink implements AudioSink {
     /** Returns a new {@link AudioTrack} for the given parameters. */
     AudioTrack getAudioTrack(
         AudioTrackConfig audioTrackConfig, AudioAttributes audioAttributes, int audioSessionId);
+
+    /** Returns the channel mask config for the given channel count. */
+    default int getAudioTrackChannelConfig(int channelCount) {
+      return Util.getAudioTrackChannelConfig(channelCount);
+    }
   }
 
   /**
@@ -731,7 +736,10 @@ public final class DefaultAudioSink implements AudioSink {
       outputMode = OUTPUT_MODE_PCM;
       outputEncoding = outputFormat.encoding;
       outputSampleRate = outputFormat.sampleRate;
-      outputChannelConfig = Util.getAudioTrackChannelConfig(outputFormat.channelCount);
+
+      outputChannelConfig =
+          audioTrackProvider.getAudioTrackChannelConfig(outputFormat.channelCount);
+
       outputPcmFrameSize = Util.getPcmFrameSize(outputEncoding, outputFormat.channelCount);
       enableAudioTrackPlaybackParams = preferAudioTrackPlaybackParams;
     } else {
@@ -748,7 +756,10 @@ public final class DefaultAudioSink implements AudioSink {
         outputMode = OUTPUT_MODE_OFFLOAD;
         outputEncoding =
             MimeTypes.getEncoding(checkNotNull(inputFormat.sampleMimeType), inputFormat.codecs);
-        outputChannelConfig = Util.getAudioTrackChannelConfig(inputFormat.channelCount);
+
+        outputChannelConfig =
+            audioTrackProvider.getAudioTrackChannelConfig(inputFormat.channelCount);
+
         // Offload requires AudioTrack playback parameters to apply speed changes quickly.
         enableAudioTrackPlaybackParams = true;
         enableOffloadGapless = audioOffloadSupport.isGaplessSupported;

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/DefaultVideoSink.java

@@ -78,12 +78,12 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
   }
 
   @Override
-  public void onStarted() {
+  public void startRendering() {
     videoFrameReleaseControl.onStarted();
   }
 
   @Override
-  public void onStopped() {
+  public void stopRendering() {
     videoFrameReleaseControl.onStopped();
   }
 

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/MediaCodecVideoRenderer.java

@@ -153,6 +153,17 @@ public class MediaCodecVideoRenderer extends MediaCodecRenderer
    */
   private static final long OFFSET_FROM_PERIOD_END_TO_TREAT_AS_LAST_US = 100_000L;
 
+  /**
+   * The offset from {@link #getLastResetPositionUs()} in microseconds, before which input buffers
+   * are not allowed to be dropped.
+   *
+   * <p>This value must be greater than the pre-roll distance used by common audio codecs, such as
+   * 80ms used by Opus <a
+   * href="https://opus-codec.org/docs/opus_in_isobmff.html#4.3.6.2">Encapsulation of Opus in ISO
+   * Base Media File Format</a>
+   */
+  private static final long OFFSET_FROM_RESET_POSITION_TO_ALLOW_INPUT_BUFFER_DROPPING_US = 200_000L;
+
   /**
    * The maximum number of consecutive dropped input buffers that allow discarding frame headers.
    *
@@ -644,7 +655,16 @@ public class MediaCodecVideoRenderer extends MediaCodecRenderer
       boolean treatDroppedBuffersAsSkipped)
       throws ExoPlaybackException {
     if (minEarlyUsToDropDecoderInput != C.TIME_UNSET) {
-      shouldDropDecoderInputBuffers = earlyUs < minEarlyUsToDropDecoderInput;
+      // TODO: b/161996553 - Remove the isAwayFromLastResetPosition check when audio pre-rolling
+      // is implemented correctly. Audio codecs such as Opus require pre-roll samples to be decoded
+      // and discarded on a seek. Depending on the audio decoder, the positionUs may jump forward
+      // by the pre-roll duration. Do not drop more frames than necessary when this happens.
+      boolean isAwayFromLastResetPosition =
+          positionUs
+              > getLastResetPositionUs()
+                  + OFFSET_FROM_RESET_POSITION_TO_ALLOW_INPUT_BUFFER_DROPPING_US;
+      shouldDropDecoderInputBuffers =
+          isAwayFromLastResetPosition && earlyUs < minEarlyUsToDropDecoderInput;
     }
     return shouldDropBuffersToKeyframe(earlyUs, elapsedRealtimeUs, isLastFrame)
         && maybeDropBuffersToKeyframe(positionUs, treatDroppedBuffersAsSkipped);
@@ -1058,7 +1078,7 @@ public class MediaCodecVideoRenderer extends MediaCodecRenderer
     totalVideoFrameProcessingOffsetUs = 0;
     videoFrameProcessingOffsetCount = 0;
     if (videoSink != null) {
-      videoSink.onStarted();
+      videoSink.startRendering();
     } else {
       videoFrameReleaseControl.onStarted();
     }
@@ -1069,7 +1089,7 @@ public class MediaCodecVideoRenderer extends MediaCodecRenderer
     maybeNotifyDroppedFrames();
     maybeNotifyVideoFrameProcessingOffset();
     if (videoSink != null) {
-      videoSink.onStopped();
+      videoSink.stopRendering();
     } else {
       videoFrameReleaseControl.onStopped();
     }

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/PlaybackVideoGraphWrapper.java

@@ -701,13 +701,13 @@ public final class PlaybackVideoGraphWrapper implements VideoSinkProvider, Video
     }
 
     @Override
-    public void onStarted() {
+    public void startRendering() {
-      defaultVideoSink.onStarted();
+      defaultVideoSink.startRendering();
     }
 
     @Override
-    public void onStopped() {
+    public void stopRendering() {
-      defaultVideoSink.onStopped();
+      defaultVideoSink.stopRendering();
     }
 
     @Override

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/video/VideoSink.java

@@ -139,11 +139,11 @@ public interface VideoSink {
    */
   int RELEASE_FIRST_FRAME_WHEN_PREVIOUS_STREAM_PROCESSED = 2;
 
-  /** Called when rendering starts. */
+  /** Starts rendering to the output surface. */
-  void onStarted();
+  void startRendering();
 
-  /** Called when rendering stops. */
+  /** Stops rendering to the output surface. */
-  void onStopped();
+  void stopRendering();
 
   /**
    * Sets a {@link Listener} on this sink. Callbacks are triggered on the supplied {@link Executor}.
@@ -263,8 +263,8 @@ public interface VideoSink {
       List<Effect> videoEffects);
 
   /**
-   * Allows the sink to release the first frame even if rendering is not {@linkplain #onStarted()
+   * Allows the sink to release the first frame even if rendering is not {@linkplain
-   * started}.
+   * #startRendering() started}.
    *
    * <p>This is used to update the {@link FirstFrameReleaseInstruction} of the {@linkplain
    * #onInputStreamChanged(int, Format, long, int, List) stream} that is currently being processed.

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/ExoPlayerTest.java

@@ -11787,6 +11787,54 @@ public final class ExoPlayerTest {
     player.release();
   }
 
+  @Test
+  public void enablingOffload_withFastReadingPeriodAdvancement_playerDoesNotSleep()
+      throws Exception {
+    FakeSleepRenderer sleepRenderer = new FakeSleepRenderer(C.TRACK_TYPE_AUDIO);
+    AtomicInteger sleepingForOffloadCounter = new AtomicInteger();
+    ExoPlayer player =
+        parameterizeTestExoPlayerBuilder(
+                new TestExoPlayerBuilder(context).setRenderers(sleepRenderer))
+            .build();
+    ExoPlayer.AudioOffloadListener listener =
+        new ExoPlayer.AudioOffloadListener() {
+          @Override
+          public void onSleepingForOffloadChanged(boolean sleepingForOffload) {
+            if (sleepingForOffload) {
+              sleepingForOffloadCounter.getAndIncrement();
+            }
+          }
+        };
+    player.addAudioOffloadListener(listener);
+    // Set a playlist of multiple, short audio-only items such that the reading period quickly
+    // advances past the playing period.
+    Timeline timeline = new FakeTimeline();
+    player.setMediaSources(
+        ImmutableList.of(
+            new FakeMediaSource(timeline, ExoPlayerTestRunner.AUDIO_FORMAT),
+            new FakeMediaSource(timeline, ExoPlayerTestRunner.AUDIO_FORMAT),
+            new FakeMediaSource(timeline, ExoPlayerTestRunner.AUDIO_FORMAT)));
+    player.setTrackSelectionParameters(
+        player
+            .getTrackSelectionParameters()
+            .buildUpon()
+            .setAudioOffloadPreferences(
+                new AudioOffloadPreferences.Builder()
+                    .setAudioOffloadMode(AudioOffloadPreferences.AUDIO_OFFLOAD_MODE_REQUIRED)
+                    .build())
+            .build());
+    player.prepare();
+    player.play();
+    advance(player).untilStartOfMediaItem(/* mediaItemIndex= */ 1);
+
+    sleepRenderer.sleepOnNextRender();
+    runUntilPlaybackState(player, Player.STATE_ENDED);
+
+    assertThat(sleepingForOffloadCounter.get()).isEqualTo(0);
+
+    player.release();
+  }
+
   @Test
   public void wakeupListenerWhileSleepingForOffload_isWokenUp_renderingResumes() throws Exception {
     FakeSleepRenderer sleepRenderer = new FakeSleepRenderer(C.TRACK_TYPE_AUDIO).sleepOnNextRender();

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/e2etest/ParseAv1SampleDependenciesPlaybackTest.java

@@ -29,6 +29,7 @@ import androidx.media3.exoplayer.DecoderCounters;
 import androidx.media3.exoplayer.DefaultRenderersFactory;
 import androidx.media3.exoplayer.ExoPlayer;
 import androidx.media3.exoplayer.Renderer;
+import androidx.media3.exoplayer.analytics.AnalyticsListener;
 import androidx.media3.exoplayer.audio.AudioRendererEventListener;
 import androidx.media3.exoplayer.mediacodec.MediaCodecAdapter;
 import androidx.media3.exoplayer.mediacodec.MediaCodecSelector;
@@ -103,6 +104,14 @@ public class ParseAv1SampleDependenciesPlaybackTest {
         new ExoPlayer.Builder(applicationContext, renderersFactory)
             .setClock(new FakeClock(/* isAutoAdvancing= */ true))
             .build();
+    player.addAnalyticsListener(
+        new AnalyticsListener() {
+          @Override
+          public void onDroppedVideoFrames(EventTime eventTime, int droppedFrames, long elapsedMs) {
+            // Input buffers near the reset position should not be dropped.
+            assertThat(eventTime.currentPlaybackPositionMs).isAtLeast(200);
+          }
+        });
     Surface surface = new Surface(new SurfaceTexture(/* texName= */ 1));
     player.setVideoSurface(surface);
     player.setMediaItem(MediaItem.fromUri(TEST_MP4_URI));
@@ -121,7 +130,7 @@ public class ParseAv1SampleDependenciesPlaybackTest {
     // Which input buffer is dropped first depends on the number of MediaCodec buffer slots.
     // This means the asserts cannot be isEqualTo.
     assertThat(decoderCounters.maxConsecutiveDroppedBufferCount).isAtMost(2);
-    assertThat(decoderCounters.droppedInputBufferCount).isAtLeast(8);
+    assertThat(decoderCounters.droppedInputBufferCount).isAtLeast(4);
   }
 
   private static final class CapturingRenderersFactoryWithLateThresholdToDropDecoderInputUs
@@ -155,7 +164,6 @@ public class ParseAv1SampleDependenciesPlaybackTest {
             /* enableDecoderFallback= */ false,
             eventHandler,
             videoRendererEventListener,
-            DefaultRenderersFactory.MAX_DROPPED_VIDEO_FRAME_COUNT_TO_NOTIFY,
             /* parseAv1SampleDependencies= */ true,
             /* lateThresholdToDropDecoderInputUs= */ -100_000_000L)
       };
@@ -173,7 +181,6 @@ public class ParseAv1SampleDependenciesPlaybackTest {
           boolean enableDecoderFallback,
           @Nullable Handler eventHandler,
           @Nullable VideoRendererEventListener eventListener,
-          int maxDroppedFramesToNotify,
           boolean parseAv1SampleDependencies,
           long lateThresholdToDropDecoderInputUs) {
         super(
@@ -184,7 +191,7 @@ public class ParseAv1SampleDependenciesPlaybackTest {
                 .setEnableDecoderFallback(enableDecoderFallback)
                 .setEventHandler(eventHandler)
                 .setEventListener(eventListener)
-                .setMaxDroppedFramesToNotify(maxDroppedFramesToNotify)
+                .setMaxDroppedFramesToNotify(1)
                 .experimentalSetParseAv1SampleDependencies(parseAv1SampleDependencies)
                 .experimentalSetLateThresholdToDropDecoderInputUs(
                     lateThresholdToDropDecoderInputUs));

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

@@ -81,13 +81,13 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
   }
 
   @Override
-  public void onStarted() {
+  public void startRendering() {
-    executeOrDelay(VideoSink::onStarted);
+    executeOrDelay(VideoSink::startRendering);
   }
 
   @Override
-  public void onStopped() {
+  public void stopRendering() {
-    executeOrDelay(VideoSink::onStopped);
+    executeOrDelay(VideoSink::stopRendering);
   }
 
   @Override

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

@@ -77,6 +77,7 @@ import androidx.media3.exoplayer.trackselection.ExoTrackSelection;
 import androidx.media3.exoplayer.upstream.Allocator;
 import androidx.media3.exoplayer.util.EventLogger;
 import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
+import androidx.media3.exoplayer.video.VideoFrameMetadataListener;
 import androidx.media3.exoplayer.video.VideoFrameReleaseControl;
 import com.google.common.collect.ImmutableList;
 import com.google.common.collect.ImmutableSet;
@@ -320,6 +321,7 @@ public final class CompositionPlayer extends SimpleBasePlayer
   private @MonotonicNonNull Composition composition;
   private @MonotonicNonNull Size videoOutputSize;
   private @MonotonicNonNull PlaybackVideoGraphWrapper playbackVideoGraphWrapper;
+  private @MonotonicNonNull VideoFrameMetadataListener pendingVideoFrameMetadatListener;
 
   private long compositionDurationUs;
   private boolean playWhenReady;
@@ -336,7 +338,6 @@ public final class CompositionPlayer extends SimpleBasePlayer
   private LivePositionSupplier positionSupplier;
   private LivePositionSupplier bufferedPositionSupplier;
   private LivePositionSupplier totalBufferedDurationSupplier;
-  private boolean isSeeking;
 
   // "this" reference for position suppliers.
   @SuppressWarnings("initialization:methodref.receiver.bound.invalid")
@@ -506,9 +507,9 @@ public final class CompositionPlayer extends SimpleBasePlayer
     playWhenReadyChangeReason = PLAY_WHEN_READY_CHANGE_REASON_USER_REQUEST;
     if (playbackState == STATE_READY) {
       if (playWhenReady) {
-        finalAudioSink.play();
+        checkStateNotNull(compositionPlayerInternal).startRendering();
       } else {
-        finalAudioSink.pause();
+        checkStateNotNull(compositionPlayerInternal).stopRendering();
       }
       for (int i = 0; i < players.size(); i++) {
         players.get(i).setPlayWhenReady(playWhenReady);
@@ -587,7 +588,9 @@ public final class CompositionPlayer extends SimpleBasePlayer
   @Override
   protected ListenableFuture<?> handleSetVolume(float volume) {
     this.volume = Util.constrainValue(volume, /* min= */ 0.0f, /* max= */ 1.0f);
-    finalAudioSink.setVolume(this.volume);
+    if (compositionPlayerInternal != null) {
+      compositionPlayerInternal.setVolume(this.volume);
+    }
     return Futures.immediateVoidFuture();
   }
 
@@ -597,7 +600,6 @@ public final class CompositionPlayer extends SimpleBasePlayer
     resetLivePositionSuppliers();
     CompositionPlayerInternal compositionPlayerInternal =
         checkStateNotNull(this.compositionPlayerInternal);
-    isSeeking = true;
     compositionPlayerInternal.startSeek(positionMs);
     for (int i = 0; i < players.size(); i++) {
       players.get(i).seekTo(positionMs);
@@ -606,6 +608,15 @@ public final class CompositionPlayer extends SimpleBasePlayer
     return Futures.immediateVoidFuture();
   }
 
+  /** Sets the {@link VideoFrameMetadataListener}. */
+  public void setVideoFrameMetadataListener(VideoFrameMetadataListener videoFrameMetadataListener) {
+    if (players.isEmpty()) {
+      pendingVideoFrameMetadatListener = videoFrameMetadataListener;
+      return;
+    }
+    players.get(0).setVideoFrameMetadataListener(videoFrameMetadataListener);
+  }
+
   // CompositionPlayerInternal.Listener methods
 
   @Override
@@ -683,22 +694,17 @@ public final class CompositionPlayer extends SimpleBasePlayer
         for (int i = 0; i < players.size(); i++) {
           players.get(i).setPlayWhenReady(false);
         }
-        if (!isSeeking) {
+        checkStateNotNull(compositionPlayerInternal).stopRendering();
-          // The finalAudioSink cannot be paused more than once. The audio pipeline pauses it during
-          // a seek, so don't pause here when seeking.
-          finalAudioSink.pause();
-        }
       }
     } else if (endedCount == players.size()) {
       playbackState = STATE_ENDED;
     } else {
       playbackState = STATE_READY;
-      isSeeking = false;
       if (oldPlaybackState != STATE_READY && playWhenReady) {
         for (int i = 0; i < players.size(); i++) {
           players.get(i).setPlayWhenReady(true);
         }
-        finalAudioSink.play();
+        checkStateNotNull(compositionPlayerInternal).startRendering();
       }
     }
   }
@@ -775,6 +781,9 @@ public final class CompositionPlayer extends SimpleBasePlayer
 
       if (i == 0) {
         setPrimaryPlayerSequence(player, editedMediaItemSequence);
+        if (pendingVideoFrameMetadatListener != null) {
+          player.setVideoFrameMetadataListener(pendingVideoFrameMetadatListener);
+        }
       } else {
         setSecondaryPlayerSequence(player, editedMediaItemSequence, primarySequenceDurationUs);
       }
@@ -799,6 +808,7 @@ public final class CompositionPlayer extends SimpleBasePlayer
             playbackVideoGraphWrapper,
             /* listener= */ this,
             compositionInternalListenerHandler);
+    compositionPlayerInternal.setVolume(volume);
   }
 
   private void setPrimaryPlayerSequence(ExoPlayer player, EditedMediaItemSequence sequence) {

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

@@ -47,11 +47,14 @@ import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
   }
 
   private static final String TAG = "CompPlayerInternal";
-  private static final int MSG_SET_OUTPUT_SURFACE_INFO = 1;
+  private static final int MSG_START_RENDERING = 1;
-  private static final int MSG_CLEAR_OUTPUT_SURFACE = 2;
+  private static final int MSG_STOP_RENDERING = 2;
-  private static final int MSG_START_SEEK = 3;
+  private static final int MSG_SET_VOLUME = 3;
-  private static final int MSG_END_SEEK = 4;
+  private static final int MSG_SET_OUTPUT_SURFACE_INFO = 4;
-  private static final int MSG_RELEASE = 5;
+  private static final int MSG_CLEAR_OUTPUT_SURFACE = 5;
+  private static final int MSG_START_SEEK = 6;
+  private static final int MSG_END_SEEK = 7;
+  private static final int MSG_RELEASE = 8;
 
   private final Clock clock;
   private final HandlerWrapper handler;
@@ -94,6 +97,18 @@ import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
 
   // Public methods
 
+  public void startRendering() {
+    handler.sendEmptyMessage(MSG_START_RENDERING);
+  }
+
+  public void stopRendering() {
+    handler.sendEmptyMessage(MSG_STOP_RENDERING);
+  }
+
+  public void setVolume(float volume) {
+    handler.obtainMessage(MSG_SET_VOLUME, volume).sendToTarget();
+  }
+
   /** Sets the output surface information on the video pipeline. */
   public void setOutputSurfaceInfo(Surface surface, Size size) {
     handler
@@ -103,7 +118,7 @@ import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
 
   /** Clears the output surface from the video pipeline. */
   public void clearOutputSurface() {
-    handler.obtainMessage(MSG_CLEAR_OUTPUT_SURFACE).sendToTarget();
+    handler.sendEmptyMessage(MSG_CLEAR_OUTPUT_SURFACE);
   }
 
   public void startSeek(long positionMs) {
@@ -111,7 +126,7 @@ import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
   }
 
   public void endSeek() {
-    handler.obtainMessage(MSG_END_SEEK).sendToTarget();
+    handler.sendEmptyMessage(MSG_END_SEEK);
   }
 
   /**
@@ -139,6 +154,15 @@ import androidx.media3.exoplayer.video.PlaybackVideoGraphWrapper;
   public boolean handleMessage(Message message) {
     try {
       switch (message.what) {
+        case MSG_START_RENDERING:
+          playbackAudioGraphWrapper.startRendering();
+          break;
+        case MSG_STOP_RENDERING:
+          playbackAudioGraphWrapper.stopRendering();
+          break;
+        case MSG_SET_VOLUME:
+          playbackAudioGraphWrapper.setVolume(/* volume= */ (float) message.obj);
+          break;
         case MSG_SET_OUTPUT_SURFACE_INFO:
           setOutputSurfaceInfoOnInternalThread(
               /* outputSurfaceInfo= */ (OutputSurfaceInfo) message.obj);

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

@@ -47,6 +47,7 @@ import java.util.Objects;
   private AudioFormat outputAudioFormat;
   private long outputFramesWritten;
   private long seekPositionUs;
+  private boolean isRenderingStarted;
 
   /**
    * Creates an instance.
@@ -135,6 +136,24 @@ import java.util.Objects;
         + sampleCountToDurationUs(outputFramesWritten, outputAudioFormat.sampleRate);
   }
 
+  public void startRendering() {
+    finalAudioSink.play();
+    isRenderingStarted = true;
+  }
+
+  public void stopRendering() {
+    if (!isRenderingStarted) {
+      // The finalAudioSink cannot be paused more than once.
+      return;
+    }
+    finalAudioSink.pause();
+    isRenderingStarted = false;
+  }
+
+  public void setVolume(float volume) {
+    finalAudioSink.setVolume(volume);
+  }
+
   /**
    * Handles the steps that need to be executed for a seek before seeking the upstream players.
    *
@@ -144,7 +163,7 @@ import java.util.Objects;
     if (positionUs == C.TIME_UNSET) {
       positionUs = 0;
     }
-    finalAudioSink.pause();
+    stopRendering();
     audioGraph.blockInput();
     audioGraph.setPendingStartTimeUs(positionUs);
     audioGraph.flush();

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

@@ -576,7 +576,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
     @Override
     protected void onStarted() throws ExoPlaybackException {
       super.onStarted();
-      videoSink.onStarted();
+      videoSink.startRendering();
     }
 
     @Override
@@ -595,7 +595,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
     @Override
     protected void onStopped() {
       super.onStopped();
-      videoSink.onStopped();
+      videoSink.stopRendering();
     }
 
     @Override

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

@@ -756,8 +756,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
     private void createEncodedSampleExporter(@C.TrackType int trackType) {
       checkState(assetLoaderInputTracker.getSampleExporter(trackType) == null);
       checkArgument(
-          trackType != TRACK_TYPE_AUDIO || !composition.sequences.get(sequenceIndex).hasGaps(),
+          !composition.sequences.get(sequenceIndex).hasGaps(), "Gaps can not be transmuxed.");
-          "Gaps can not be transmuxed.");
       assetLoaderInputTracker.registerSampleExporter(
           trackType,
           new EncodedSampleExporter(

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

@@ -18,7 +18,6 @@ package androidx.media3.transformer;
 
 import static androidx.media3.common.ColorInfo.SDR_BT709_LIMITED;
 import static androidx.media3.common.ColorInfo.isTransferHdr;
-import static androidx.media3.common.util.Assertions.checkArgument;
 import static androidx.media3.exoplayer.mediacodec.MediaCodecUtil.getAlternativeCodecMimeType;
 import static androidx.media3.transformer.Composition.HDR_MODE_KEEP_HDR;
 import static androidx.media3.transformer.Composition.HDR_MODE_TONE_MAP_HDR_TO_SDR_USING_OPEN_GL;
@@ -91,8 +90,6 @@ public final class TransformerUtil {
       MuxerWrapper muxerWrapper) {
     if (composition.sequences.size() > 1
         || composition.sequences.get(sequenceIndex).editedMediaItems.size() > 1) {
-      checkArgument(
-          !composition.hasGaps() || !composition.transmuxAudio, "Gaps can not be transmuxed.");
       return !composition.transmuxAudio;
     }
     if (composition.hasGaps()) {

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

@@ -40,11 +40,11 @@ public class BufferingVideoSinkTest {
     VideoSink videoSinkMock = mock(VideoSink.class);
 
     bufferingVideoSink.setVideoSink(videoSinkMock);
-    bufferingVideoSink.onStarted();
+    bufferingVideoSink.startRendering();
     bufferingVideoSink.flush(/* resetPosition= */ true);
 
     InOrder inOrder = Mockito.inOrder(videoSinkMock);
-    inOrder.verify(videoSinkMock).onStarted();
+    inOrder.verify(videoSinkMock).startRendering();
     inOrder.verify(videoSinkMock).flush(/* resetPosition= */ true);
   }
 
@@ -52,12 +52,12 @@ public class BufferingVideoSinkTest {
   public void setVideoSink_executesPendingOperations() {
     BufferingVideoSink bufferingVideoSink = new BufferingVideoSink(context);
     VideoSink videoSinkMock = mock(VideoSink.class);
-    bufferingVideoSink.onStarted();
+    bufferingVideoSink.startRendering();
     bufferingVideoSink.flush(/* resetPosition= */ true);
     bufferingVideoSink.setVideoSink(videoSinkMock);
 
     InOrder inOrder = Mockito.inOrder(videoSinkMock);
-    inOrder.verify(videoSinkMock).onStarted();
+    inOrder.verify(videoSinkMock).startRendering();
     inOrder.verify(videoSinkMock).flush(/* resetPosition= */ true);
   }
 
@@ -68,10 +68,10 @@ public class BufferingVideoSinkTest {
     bufferingVideoSink.setVideoSink(videoSinkMock);
 
     bufferingVideoSink.setVideoSink(null);
-    bufferingVideoSink.onStarted();
+    bufferingVideoSink.startRendering();
     bufferingVideoSink.flush(/* resetPosition= */ true);
 
-    verify(videoSinkMock, never()).onStarted();
+    verify(videoSinkMock, never()).startRendering();
     verify(videoSinkMock, never()).flush(/* resetPosition= */ true);
   }
 
@@ -80,12 +80,12 @@ public class BufferingVideoSinkTest {
     BufferingVideoSink bufferingVideoSink = new BufferingVideoSink(context);
     VideoSink videoSinkMock = mock(VideoSink.class);
 
-    bufferingVideoSink.onStarted();
+    bufferingVideoSink.startRendering();
     bufferingVideoSink.flush(/* resetPosition= */ true);
     bufferingVideoSink.clearPendingOperations();
     bufferingVideoSink.setVideoSink(videoSinkMock);
 
-    verify(videoSinkMock, never()).onStarted();
+    verify(videoSinkMock, never()).startRendering();
     verify(videoSinkMock, never()).flush(/* resetPosition= */ true);
   }
 }