Use SpeedChangingAudioProcessor in Transformer

PiperOrigin-RevId: 426113559

libraries/test_data/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump

@@ -132,148 +132,154 @@ sample:
   presentationTimeUs = 0
 sample:
   trackIndex = 0
-  dataHashCode = 1000136444
-  size = 140
+  dataHashCode = -1948569090
+  size = 72
   isKeyFrame = true
   presentationTimeUs = 417
 sample:
   trackIndex = 0
-  dataHashCode = 217961709
-  size = 172
-  isKeyFrame = true
-  presentationTimeUs = 3334
-sample:
-  trackIndex = 0
-  dataHashCode = -879376936
-  size = 176
-  isKeyFrame = true
-  presentationTimeUs = 6917
-sample:
-  trackIndex = 0
-  dataHashCode = 1259979587
-  size = 192
-  isKeyFrame = true
-  presentationTimeUs = 10584
-sample:
-  trackIndex = 0
-  dataHashCode = 907407225
-  size = 188
-  isKeyFrame = true
-  presentationTimeUs = 14584
-sample:
-  trackIndex = 0
-  dataHashCode = -904354707
-  size = 176
-  isKeyFrame = true
-  presentationTimeUs = 18500
-sample:
-  trackIndex = 0
-  dataHashCode = 1001385853
-  size = 172
-  isKeyFrame = true
-  presentationTimeUs = 22167
-sample:
-  trackIndex = 0
-  dataHashCode = 1545716086
-  size = 196
-  isKeyFrame = true
-  presentationTimeUs = 25750
-sample:
-  trackIndex = 0
-  dataHashCode = 358710839
-  size = 180
-  isKeyFrame = true
-  presentationTimeUs = 29834
-sample:
-  trackIndex = 0
-  dataHashCode = -671124798
-  size = 140
-  isKeyFrame = true
-  presentationTimeUs = 33584
-sample:
-  trackIndex = 0
-  dataHashCode = -945404910
-  size = 120
-  isKeyFrame = true
-  presentationTimeUs = 36500
-sample:
-  trackIndex = 0
-  dataHashCode = 1881048379
-  size = 88
-  isKeyFrame = true
-  presentationTimeUs = 39000
-sample:
-  trackIndex = 0
-  dataHashCode = 1059579897
-  size = 88
-  isKeyFrame = true
-  presentationTimeUs = 40834
-sample:
-  trackIndex = 0
-  dataHashCode = 1496098648
+  dataHashCode = -1316750072
   size = 84
   isKeyFrame = true
-  presentationTimeUs = 42667
+  presentationTimeUs = 1917
 sample:
   trackIndex = 0
-  dataHashCode = 250093960
-  size = 751
+  dataHashCode = 1016428949
+  size = 88
   isKeyFrame = true
-  presentationTimeUs = 44417
+  presentationTimeUs = 3667
 sample:
   trackIndex = 0
-  dataHashCode = 1895536226
-  size = 1045
+  dataHashCode = -1127325245
+  size = 96
   isKeyFrame = true
-  presentationTimeUs = 60063
+  presentationTimeUs = 5500
 sample:
   trackIndex = 0
-  dataHashCode = 1723596464
-  size = 947
+  dataHashCode = 1148147726
+  size = 92
   isKeyFrame = true
-  presentationTimeUs = 81834
+  presentationTimeUs = 7500
 sample:
   trackIndex = 0
-  dataHashCode = -978803114
-  size = 946
+  dataHashCode = -2125685540
+  size = 76
   isKeyFrame = true
-  presentationTimeUs = 101563
+  presentationTimeUs = 9417
 sample:
   trackIndex = 0
-  dataHashCode = 387377078
-  size = 946
+  dataHashCode = 473329679
+  size = 24
   isKeyFrame = true
-  presentationTimeUs = 121271
+  presentationTimeUs = 11000
 sample:
   trackIndex = 0
-  dataHashCode = -132658698
-  size = 901
+  dataHashCode = 240990900
+  size = 176
   isKeyFrame = true
-  presentationTimeUs = 140980
+  presentationTimeUs = 11500
 sample:
   trackIndex = 0
-  dataHashCode = 1495036471
-  size = 899
+  dataHashCode = 777637182
+  size = 196
   isKeyFrame = true
-  presentationTimeUs = 159750
+  presentationTimeUs = 15167
 sample:
   trackIndex = 0
-  dataHashCode = 304440590
-  size = 878
+  dataHashCode = 1872106264
+  size = 180
   isKeyFrame = true
-  presentationTimeUs = 178480
+  presentationTimeUs = 19250
 sample:
   trackIndex = 0
-  dataHashCode = -1955900344
-  size = 112
+  dataHashCode = -1520711499
+  size = 140
   isKeyFrame = true
-  presentationTimeUs = 196771
+  presentationTimeUs = 23000
 sample:
   trackIndex = 0
-  dataHashCode = 88896626
-  size = 116
+  dataHashCode = 1580199067
+  size = 232
   isKeyFrame = true
-  presentationTimeUs = 199105
+  presentationTimeUs = 25917
+sample:
+  trackIndex = 0
+  dataHashCode = 475464086
+  size = 184
+  isKeyFrame = true
+  presentationTimeUs = 30750
+sample:
+  trackIndex = 0
+  dataHashCode = -211754132
+  size = 172
+  isKeyFrame = true
+  presentationTimeUs = 34584
+sample:
+  trackIndex = 0
+  dataHashCode = 1236547164
+  size = 172
+  isKeyFrame = true
+  presentationTimeUs = 38167
+sample:
+  trackIndex = 0
+  dataHashCode = -2064216186
+  size = 188
+  isKeyFrame = true
+  presentationTimeUs = 41750
+sample:
+  trackIndex = 0
+  dataHashCode = -682950885
+  size = 260
+  isKeyFrame = true
+  presentationTimeUs = 45667
+sample:
+  trackIndex = 0
+  dataHashCode = 1301206627
+  size = 236
+  isKeyFrame = true
+  presentationTimeUs = 51084
+sample:
+  trackIndex = 0
+  dataHashCode = 256580525
+  size = 236
+  isKeyFrame = true
+  presentationTimeUs = 56000
+sample:
+  trackIndex = 0
+  dataHashCode = -1086601304
+  size = 236
+  isKeyFrame = true
+  presentationTimeUs = 60917
+sample:
+  trackIndex = 0
+  dataHashCode = -2046131588
+  size = 224
+  isKeyFrame = true
+  presentationTimeUs = 65834
+sample:
+  trackIndex = 0
+  dataHashCode = 1550955865
+  size = 224
+  isKeyFrame = true
+  presentationTimeUs = 70500
+sample:
+  trackIndex = 0
+  dataHashCode = -274800552
+  size = 220
+  isKeyFrame = true
+  presentationTimeUs = 75167
+sample:
+  trackIndex = 0
+  dataHashCode = 382420909
+  size = 224
+  isKeyFrame = true
+  presentationTimeUs = 79750
+sample:
+  trackIndex = 0
+  dataHashCode = -1431575865
+  size = 232
+  isKeyFrame = true
+  presentationTimeUs = 84417
 sample:
   trackIndex = 1
   dataHashCode = -968901399

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

@@ -20,7 +20,6 @@ import static androidx.media3.common.util.Assertions.checkNotNull;
 import static androidx.media3.common.util.Assertions.checkState;
 import static java.lang.Math.min;
 
-import android.media.MediaCodec.BufferInfo;
 import androidx.annotation.Nullable;
 import androidx.media3.common.C;
 import androidx.media3.common.Format;
@@ -28,9 +27,9 @@ import androidx.media3.common.util.Util;
 import androidx.media3.decoder.DecoderInputBuffer;
 import androidx.media3.exoplayer.audio.AudioProcessor;
 import androidx.media3.exoplayer.audio.AudioProcessor.AudioFormat;
-import androidx.media3.exoplayer.audio.SonicAudioProcessor;
 import java.nio.ByteBuffer;
 import java.util.List;
+import org.checkerframework.checker.nullness.qual.RequiresNonNull;
 import org.checkerframework.dataflow.qual.Pure;
 
 /**
@@ -43,22 +42,18 @@ import org.checkerframework.dataflow.qual.Pure;
   private final Codec decoder;
   private final DecoderInputBuffer decoderInputBuffer;
 
-  private final SonicAudioProcessor sonicAudioProcessor;
-  private final SpeedProvider speedProvider;
-  private final boolean flattenForSlowMotion;
+  @Nullable private final SpeedChangingAudioProcessor speedChangingAudioProcessor;
 
   private final Codec encoder;
   private final AudioFormat encoderInputAudioFormat;
   private final DecoderInputBuffer encoderInputBuffer;
   private final DecoderInputBuffer encoderOutputBuffer;
 
+  private ByteBuffer processorOutputBuffer;
+
   private long nextEncoderInputBufferTimeUs;
   private long encoderBufferDurationRemainder;
 
-  private ByteBuffer sonicOutputBuffer;
-  private boolean drainingSonicForSpeedChange;
-  private float currentSpeed;
-
   public AudioTranscodingSamplePipeline(
       Format inputFormat,
       TransformationRequest transformationRequest,
@@ -74,13 +69,8 @@ import org.checkerframework.dataflow.qual.Pure;
     encoderOutputBuffer =
         new DecoderInputBuffer(DecoderInputBuffer.BUFFER_REPLACEMENT_MODE_DISABLED);
 
-    this.decoder = decoderFactory.createForAudioDecoding(inputFormat);
+    decoder = decoderFactory.createForAudioDecoding(inputFormat);
 
-    this.flattenForSlowMotion = transformationRequest.flattenForSlowMotion;
-    sonicAudioProcessor = new SonicAudioProcessor();
-    sonicOutputBuffer = AudioProcessor.EMPTY_BUFFER;
-    speedProvider = new SegmentSpeedProvider(inputFormat);
-    currentSpeed = speedProvider.getSpeed(0);
     AudioFormat encoderInputAudioFormat =
         new AudioFormat(
             inputFormat.sampleRate,
@@ -88,18 +78,21 @@ import org.checkerframework.dataflow.qual.Pure;
             // The decoder uses ENCODING_PCM_16BIT by default.
             // https://developer.android.com/reference/android/media/MediaCodec#raw-audio-buffers
             C.ENCODING_PCM_16BIT);
-    if (flattenForSlowMotion) {
+    if (transformationRequest.flattenForSlowMotion) {
+      speedChangingAudioProcessor =
+          new SpeedChangingAudioProcessor(new SegmentSpeedProvider(inputFormat));
       try {
-        encoderInputAudioFormat = sonicAudioProcessor.configure(encoderInputAudioFormat);
+        encoderInputAudioFormat = speedChangingAudioProcessor.configure(encoderInputAudioFormat);
       } catch (AudioProcessor.UnhandledAudioFormatException impossible) {
         throw new IllegalStateException(impossible);
       }
-      sonicAudioProcessor.setSpeed(currentSpeed);
-      sonicAudioProcessor.setPitch(currentSpeed);
-      sonicAudioProcessor.flush();
+      speedChangingAudioProcessor.flush();
+    } else {
+      speedChangingAudioProcessor = null;
     }
-    this.encoderInputAudioFormat = encoderInputAudioFormat;
+    processorOutputBuffer = AudioProcessor.EMPTY_BUFFER;
 
+    this.encoderInputAudioFormat = encoderInputAudioFormat;
     Format requestedOutputFormat =
         new Format.Builder()
             .setSampleMimeType(
@@ -130,8 +123,8 @@ import org.checkerframework.dataflow.qual.Pure;
 
   @Override
   public boolean processData() throws TransformationException {
-    if (sonicAudioProcessor.isActive()) {
-      return feedEncoderFromSonic() || feedSonicFromDecoder();
+    if (speedChangingAudioProcessor != null) {
+      return feedEncoderFromProcessor() || feedProcessorFromDecoder();
     } else {
       return feedEncoderFromDecoder();
     }
@@ -167,7 +160,9 @@ import org.checkerframework.dataflow.qual.Pure;
 
   @Override
   public void release() {
-    sonicAudioProcessor.reset();
+    if (speedChangingAudioProcessor != null) {
+      speedChangingAudioProcessor.reset();
+    }
     decoder.release();
     encoder.release();
   }
@@ -190,10 +185,7 @@ import org.checkerframework.dataflow.qual.Pure;
     if (decoderOutputBuffer == null) {
       return false;
     }
-    if (isSpeedChanging(checkNotNull(decoder.getOutputBufferInfo()))) {
-      flushSonicAndSetSpeed(currentSpeed);
-      return false;
-    }
+
     feedEncoder(decoderOutputBuffer);
     if (!decoderOutputBuffer.hasRemaining()) {
       decoder.releaseOutputBuffer();
@@ -205,22 +197,23 @@ import org.checkerframework.dataflow.qual.Pure;
    * Attempts to pass audio processor output data to the encoder, and returns whether it may be
    * possible to pass more data immediately by calling this method again.
    */
-  private boolean feedEncoderFromSonic() throws TransformationException {
+  @RequiresNonNull("speedChangingAudioProcessor")
+  private boolean feedEncoderFromProcessor() throws TransformationException {
     if (!encoder.maybeDequeueInputBuffer(encoderInputBuffer)) {
       return false;
     }
 
-    if (!sonicOutputBuffer.hasRemaining()) {
-      sonicOutputBuffer = sonicAudioProcessor.getOutput();
-      if (!sonicOutputBuffer.hasRemaining()) {
-        if (decoder.isEnded() && sonicAudioProcessor.isEnded()) {
+    if (!processorOutputBuffer.hasRemaining()) {
+      processorOutputBuffer = speedChangingAudioProcessor.getOutput();
+      if (!processorOutputBuffer.hasRemaining()) {
+        if (decoder.isEnded() && speedChangingAudioProcessor.isEnded()) {
           queueEndOfStreamToEncoder();
         }
         return false;
       }
     }
 
-    feedEncoder(sonicOutputBuffer);
+    feedEncoder(processorOutputBuffer);
     return true;
   }
 
@@ -228,37 +221,27 @@ import org.checkerframework.dataflow.qual.Pure;
    * Attempts to process decoder output data, and returns whether it may be possible to process more
    * data immediately by calling this method again.
    */
-  private boolean feedSonicFromDecoder() throws TransformationException {
-    if (drainingSonicForSpeedChange) {
-      if (sonicAudioProcessor.isEnded() && !sonicOutputBuffer.hasRemaining()) {
-        flushSonicAndSetSpeed(currentSpeed);
-        drainingSonicForSpeedChange = false;
-      }
-      return false;
-    }
-
-    // Sonic invalidates any previous output buffer when more input is queued, so we don't queue if
-    // there is output still to be processed.
-    if (sonicOutputBuffer.hasRemaining()) {
+  @RequiresNonNull("speedChangingAudioProcessor")
+  private boolean feedProcessorFromDecoder() throws TransformationException {
+    // Audio processors invalidate any previous output buffer when more input is queued, so we don't
+    // queue if there is output still to be processed.
+    if (processorOutputBuffer.hasRemaining()
+        || speedChangingAudioProcessor.getOutput().hasRemaining()) {
       return false;
     }
 
     if (decoder.isEnded()) {
-      sonicAudioProcessor.queueEndOfStream();
+      speedChangingAudioProcessor.queueEndOfStream();
       return false;
     }
-    checkState(!sonicAudioProcessor.isEnded());
+    checkState(!speedChangingAudioProcessor.isEnded());
 
     @Nullable ByteBuffer decoderOutputBuffer = decoder.getOutputBuffer();
     if (decoderOutputBuffer == null) {
       return false;
     }
-    if (isSpeedChanging(checkNotNull(decoder.getOutputBufferInfo()))) {
-      sonicAudioProcessor.queueEndOfStream();
-      drainingSonicForSpeedChange = true;
-      return false;
-    }
-    sonicAudioProcessor.queueInput(decoderOutputBuffer);
+
+    speedChangingAudioProcessor.queueInput(decoderOutputBuffer);
     if (!decoderOutputBuffer.hasRemaining()) {
       decoder.releaseOutputBuffer();
     }
@@ -294,22 +277,6 @@ import org.checkerframework.dataflow.qual.Pure;
     encoder.queueInputBuffer(encoderInputBuffer);
   }
 
-  private boolean isSpeedChanging(BufferInfo bufferInfo) {
-    if (!flattenForSlowMotion) {
-      return false;
-    }
-    float newSpeed = speedProvider.getSpeed(bufferInfo.presentationTimeUs);
-    boolean speedChanging = newSpeed != currentSpeed;
-    currentSpeed = newSpeed;
-    return speedChanging;
-  }
-
-  private void flushSonicAndSetSpeed(float speed) {
-    sonicAudioProcessor.setSpeed(speed);
-    sonicAudioProcessor.setPitch(speed);
-    sonicAudioProcessor.flush();
-  }
-
   private void computeNextEncoderInputBufferTimeUs(
       long bytesWritten, int bytesPerFrame, int sampleRate) {
     // The calculation below accounts for remainders and rounding. Without that it corresponds to

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

@@ -28,6 +28,8 @@ import java.nio.ByteBuffer;
 /**
  * An {@link AudioProcessor} that changes the speed of audio samples depending on their timestamp.
  */
+// TODO(b/198772621): Consider making the processor inactive and skipping it in the processor chain
+//  when speed is 1.
 /* package */ final class SpeedChangingAudioProcessor extends BaseAudioProcessor {
 
   /** The speed provider that provides the speed for each timestamp. */

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

@@ -30,7 +30,6 @@ import androidx.media3.common.Format;
 import androidx.media3.common.util.Util;
 import androidx.media3.decoder.DecoderInputBuffer;
 import java.util.List;
-import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
 import org.checkerframework.dataflow.qual.Pure;
 
 /**
@@ -42,11 +41,11 @@ import org.checkerframework.dataflow.qual.Pure;
   private final DecoderInputBuffer decoderInputBuffer;
   private final Codec decoder;
 
+  @Nullable private final FrameEditor frameEditor;
+
   private final Codec encoder;
   private final DecoderInputBuffer encoderOutputBuffer;
 
-  private @MonotonicNonNull FrameEditor frameEditor;
-
   private boolean waitingForFrameEditorInput;
 
   public VideoTranscodingSamplePipeline(
@@ -139,6 +138,8 @@ import org.checkerframework.dataflow.qual.Pure;
               /* outputSurface= */ checkNotNull(encoder.getInputSurface()),
               transformationRequest.enableHdrEditing,
               debugViewProvider);
+    } else {
+      frameEditor = null;
     }
 
     decoder =