diff --git a/libraries/test_data/src/test/assets/transformerdumps/mkv/sample_with_srt.mkv.dump b/libraries/test_data/src/test/assets/transformerdumps/mkv/sample_with_srt.mkv.dump
index 4709eeda17..5789404820 100644
--- a/libraries/test_data/src/test/assets/transformerdumps/mkv/sample_with_srt.mkv.dump
+++ b/libraries/test_data/src/test/assets/transformerdumps/mkv/sample_with_srt.mkv.dump
@@ -14,6 +14,72 @@ format 1:
   initializationData:
     data = length 30, hash F6F3D010
     data = length 10, hash 7A0D0F2B
+sample:
+  trackIndex = 0
+  dataHashCode = 555688582
+  size = 416
+  isKeyFrame = true
+  presentationTimeUs = 0
+sample:
+  trackIndex = 0
+  dataHashCode = 2000837254
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 4717
+sample:
+  trackIndex = 0
+  dataHashCode = -1593942879
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 9456
+sample:
+  trackIndex = 0
+  dataHashCode = 587837542
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 14196
+sample:
+  trackIndex = 0
+  dataHashCode = -1836423877
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 18935
+sample:
+  trackIndex = 0
+  dataHashCode = 874705099
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 23674
+sample:
+  trackIndex = 0
+  dataHashCode = -269206181
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 28413
+sample:
+  trackIndex = 0
+  dataHashCode = -58682425
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 33152
+sample:
+  trackIndex = 0
+  dataHashCode = -859796970
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 37892
+sample:
+  trackIndex = 0
+  dataHashCode = 711911523
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 42631
+sample:
+  trackIndex = 0
+  dataHashCode = -694513071
+  size = 418
+  isKeyFrame = true
+  presentationTimeUs = 47370
 sample:
   trackIndex = 1
   dataHashCode = -252482306
@@ -104,72 +170,6 @@ sample:
   size = 5525
   isKeyFrame = false
   presentationTimeUs = 567000
-sample:
-  trackIndex = 0
-  dataHashCode = 555688582
-  size = 416
-  isKeyFrame = true
-  presentationTimeUs = 0
-sample:
-  trackIndex = 0
-  dataHashCode = 2000837254
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 4717
-sample:
-  trackIndex = 0
-  dataHashCode = -1593942879
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 9456
-sample:
-  trackIndex = 0
-  dataHashCode = 587837542
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 14196
-sample:
-  trackIndex = 0
-  dataHashCode = -1836423877
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 18935
-sample:
-  trackIndex = 0
-  dataHashCode = 874705099
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 23674
-sample:
-  trackIndex = 0
-  dataHashCode = -269206181
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 28413
-sample:
-  trackIndex = 0
-  dataHashCode = -58682425
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 33152
-sample:
-  trackIndex = 0
-  dataHashCode = -859796970
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 37892
-sample:
-  trackIndex = 0
-  dataHashCode = 711911523
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 42631
-sample:
-  trackIndex = 0
-  dataHashCode = -694513071
-  size = 418
-  isKeyFrame = true
-  presentationTimeUs = 47370
 sample:
   trackIndex = 0
   dataHashCode = -1124371059
@@ -200,6 +200,12 @@ sample:
   size = 418
   isKeyFrame = true
   presentationTimeUs = 71066
+sample:
+  trackIndex = 1
+  dataHashCode = -637074022
+  size = 1082
+  isKeyFrame = false
+  presentationTimeUs = 500000
 sample:
   trackIndex = 0
   dataHashCode = 1297086772
@@ -278,12 +284,6 @@ sample:
   size = 418
   isKeyFrame = true
   presentationTimeUs = 132676
-sample:
-  trackIndex = 1
-  dataHashCode = -637074022
-  size = 1082
-  isKeyFrame = false
-  presentationTimeUs = 500000
 sample:
   trackIndex = 1
   dataHashCode = -1824027029
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.48000hz.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.48000hz.dump
index b11eb46460..80419c61a0 100644
--- a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.48000hz.dump
+++ b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.48000hz.dump
@@ -14,6 +14,78 @@ format 1:
   initializationData:
     data = length 29, hash 4746B5D9
     data = length 10, hash 7A0D0F2B
+sample:
+  trackIndex = 0
+  dataHashCode = 1868041800
+  size = 22
+  isKeyFrame = true
+  presentationTimeUs = 0
+sample:
+  trackIndex = 0
+  dataHashCode = 1552866193
+  size = 8
+  isKeyFrame = true
+  presentationTimeUs = 230
+sample:
+  trackIndex = 0
+  dataHashCode = 701848493
+  size = 160
+  isKeyFrame = true
+  presentationTimeUs = 313
+sample:
+  trackIndex = 0
+  dataHashCode = 2000505720
+  size = 206
+  isKeyFrame = true
+  presentationTimeUs = 1980
+sample:
+  trackIndex = 0
+  dataHashCode = 57276327
+  size = 222
+  isKeyFrame = true
+  presentationTimeUs = 4125
+sample:
+  trackIndex = 0
+  dataHashCode = 382237408
+  size = 228
+  isKeyFrame = true
+  presentationTimeUs = 6438
+sample:
+  trackIndex = 0
+  dataHashCode = 795308374
+  size = 228
+  isKeyFrame = true
+  presentationTimeUs = 8813
+sample:
+  trackIndex = 0
+  dataHashCode = 356270426
+  size = 224
+  isKeyFrame = true
+  presentationTimeUs = 11188
+sample:
+  trackIndex = 0
+  dataHashCode = 963641614
+  size = 244
+  isKeyFrame = true
+  presentationTimeUs = 13521
+sample:
+  trackIndex = 0
+  dataHashCode = 2045272193
+  size = 234
+  isKeyFrame = true
+  presentationTimeUs = 16063
+sample:
+  trackIndex = 0
+  dataHashCode = -251933090
+  size = 228
+  isKeyFrame = true
+  presentationTimeUs = 18500
+sample:
+  trackIndex = 0
+  dataHashCode = -329315280
+  size = 234
+  isKeyFrame = true
+  presentationTimeUs = 20875
 sample:
   trackIndex = 1
   dataHashCode = -770308242
@@ -104,78 +176,6 @@ sample:
   size = 5450
   isKeyFrame = false
   presentationTimeUs = 567233
-sample:
-  trackIndex = 0
-  dataHashCode = 1868041800
-  size = 22
-  isKeyFrame = true
-  presentationTimeUs = 0
-sample:
-  trackIndex = 0
-  dataHashCode = 1552866193
-  size = 8
-  isKeyFrame = true
-  presentationTimeUs = 230
-sample:
-  trackIndex = 0
-  dataHashCode = 701848493
-  size = 160
-  isKeyFrame = true
-  presentationTimeUs = 313
-sample:
-  trackIndex = 0
-  dataHashCode = 2000505720
-  size = 206
-  isKeyFrame = true
-  presentationTimeUs = 1980
-sample:
-  trackIndex = 0
-  dataHashCode = 57276327
-  size = 222
-  isKeyFrame = true
-  presentationTimeUs = 4125
-sample:
-  trackIndex = 0
-  dataHashCode = 382237408
-  size = 228
-  isKeyFrame = true
-  presentationTimeUs = 6438
-sample:
-  trackIndex = 0
-  dataHashCode = 795308374
-  size = 228
-  isKeyFrame = true
-  presentationTimeUs = 8813
-sample:
-  trackIndex = 0
-  dataHashCode = 356270426
-  size = 224
-  isKeyFrame = true
-  presentationTimeUs = 11188
-sample:
-  trackIndex = 0
-  dataHashCode = 963641614
-  size = 244
-  isKeyFrame = true
-  presentationTimeUs = 13521
-sample:
-  trackIndex = 0
-  dataHashCode = 2045272193
-  size = 234
-  isKeyFrame = true
-  presentationTimeUs = 16063
-sample:
-  trackIndex = 0
-  dataHashCode = -251933090
-  size = 228
-  isKeyFrame = true
-  presentationTimeUs = 18500
-sample:
-  trackIndex = 0
-  dataHashCode = -329315280
-  size = 234
-  isKeyFrame = true
-  presentationTimeUs = 20875
 sample:
   trackIndex = 0
   dataHashCode = -1536715689
@@ -284,6 +284,12 @@ sample:
   size = 250
   isKeyFrame = true
   presentationTimeUs = 67625
+sample:
+  trackIndex = 1
+  dataHashCode = -1830836678
+  size = 1051
+  isKeyFrame = false
+  presentationTimeUs = 500500
 sample:
   trackIndex = 0
   dataHashCode = -1290952882
@@ -374,12 +380,6 @@ sample:
   size = 8
   isKeyFrame = true
   presentationTimeUs = 107646
-sample:
-  trackIndex = 1
-  dataHashCode = -1830836678
-  size = 1051
-  isKeyFrame = false
-  presentationTimeUs = 500500
 sample:
   trackIndex = 1
   dataHashCode = 1767407540
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.dump
index 2ee74557d5..19def158f3 100644
--- a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.dump
+++ b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample.mp4.dump
@@ -21,6 +21,12 @@ format 1:
   initializationData:
     data = length 29, hash 4746B5D9
     data = length 10, hash 7A0D0F2B
+sample:
+  trackIndex = 0
+  dataHashCode = 1205768497
+  size = 23
+  isKeyFrame = true
+  presentationTimeUs = 44000
 sample:
   trackIndex = 1
   dataHashCode = -770308242
@@ -111,12 +117,6 @@ sample:
   size = 5450
   isKeyFrame = false
   presentationTimeUs = 567233
-sample:
-  trackIndex = 0
-  dataHashCode = 1205768497
-  size = 23
-  isKeyFrame = true
-  presentationTimeUs = 44000
 sample:
   trackIndex = 0
   dataHashCode = 837571078
@@ -129,6 +129,12 @@ sample:
   size = 148
   isKeyFrame = true
   presentationTimeUs = 90439
+sample:
+  trackIndex = 1
+  dataHashCode = -1830836678
+  size = 1051
+  isKeyFrame = false
+  presentationTimeUs = 500500
 sample:
   trackIndex = 0
   dataHashCode = -822987359
@@ -238,17 +244,41 @@ sample:
   isKeyFrame = true
   presentationTimeUs = 508399
 sample:
-  trackIndex = 1
-  dataHashCode = -1830836678
-  size = 1051
-  isKeyFrame = false
-  presentationTimeUs = 500500
+  trackIndex = 0
+  dataHashCode = 1675188486
+  size = 228
+  isKeyFrame = true
+  presentationTimeUs = 531619
+sample:
+  trackIndex = 0
+  dataHashCode = 888567545
+  size = 238
+  isKeyFrame = true
+  presentationTimeUs = 554839
+sample:
+  trackIndex = 0
+  dataHashCode = -439631803
+  size = 234
+  isKeyFrame = true
+  presentationTimeUs = 578058
+sample:
+  trackIndex = 0
+  dataHashCode = 1606694497
+  size = 231
+  isKeyFrame = true
+  presentationTimeUs = 601278
 sample:
   trackIndex = 1
   dataHashCode = 1767407540
   size = 874
   isKeyFrame = false
   presentationTimeUs = 467133
+sample:
+  trackIndex = 0
+  dataHashCode = 1747388653
+  size = 217
+  isKeyFrame = true
+  presentationTimeUs = 624498
 sample:
   trackIndex = 1
   dataHashCode = 918440283
@@ -327,36 +357,6 @@ sample:
   size = 568
   isKeyFrame = false
   presentationTimeUs = 934266
-sample:
-  trackIndex = 0
-  dataHashCode = 1675188486
-  size = 228
-  isKeyFrame = true
-  presentationTimeUs = 531619
-sample:
-  trackIndex = 0
-  dataHashCode = 888567545
-  size = 238
-  isKeyFrame = true
-  presentationTimeUs = 554839
-sample:
-  trackIndex = 0
-  dataHashCode = -439631803
-  size = 234
-  isKeyFrame = true
-  presentationTimeUs = 578058
-sample:
-  trackIndex = 0
-  dataHashCode = 1606694497
-  size = 231
-  isKeyFrame = true
-  presentationTimeUs = 601278
-sample:
-  trackIndex = 0
-  dataHashCode = 1747388653
-  size = 217
-  isKeyFrame = true
-  presentationTimeUs = 624498
 sample:
   trackIndex = 0
   dataHashCode = -734560004
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump
index fc0a46dd7d..bdd2160cc1 100644
--- a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump
+++ b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump
@@ -15,6 +15,78 @@ format 1:
   initializationData:
     data = length 33, hash D3FB879D
     data = length 10, hash 7A0D0F2B
+sample:
+  trackIndex = 0
+  dataHashCode = -212376212
+  size = 20
+  isKeyFrame = true
+  presentationTimeUs = 0
+sample:
+  trackIndex = 0
+  dataHashCode = -1948569090
+  size = 72
+  isKeyFrame = true
+  presentationTimeUs = 417
+sample:
+  trackIndex = 0
+  dataHashCode = -1316750072
+  size = 84
+  isKeyFrame = true
+  presentationTimeUs = 1917
+sample:
+  trackIndex = 0
+  dataHashCode = 1016428949
+  size = 88
+  isKeyFrame = true
+  presentationTimeUs = 3667
+sample:
+  trackIndex = 0
+  dataHashCode = -1127325245
+  size = 96
+  isKeyFrame = true
+  presentationTimeUs = 5500
+sample:
+  trackIndex = 0
+  dataHashCode = 1148147726
+  size = 92
+  isKeyFrame = true
+  presentationTimeUs = 7500
+sample:
+  trackIndex = 0
+  dataHashCode = -2125685540
+  size = 76
+  isKeyFrame = true
+  presentationTimeUs = 9417
+sample:
+  trackIndex = 0
+  dataHashCode = 473329679
+  size = 24
+  isKeyFrame = true
+  presentationTimeUs = 11000
+sample:
+  trackIndex = 0
+  dataHashCode = 240990900
+  size = 176
+  isKeyFrame = true
+  presentationTimeUs = 11500
+sample:
+  trackIndex = 0
+  dataHashCode = 777637182
+  size = 196
+  isKeyFrame = true
+  presentationTimeUs = 15167
+sample:
+  trackIndex = 0
+  dataHashCode = 1872106264
+  size = 180
+  isKeyFrame = true
+  presentationTimeUs = 19250
+sample:
+  trackIndex = 0
+  dataHashCode = -1520711499
+  size = 140
+  isKeyFrame = true
+  presentationTimeUs = 23000
 sample:
   trackIndex = 1
   dataHashCode = 1949079733
@@ -123,78 +195,6 @@ sample:
   size = 1109
   isKeyFrame = false
   presentationTimeUs = 600750
-sample:
-  trackIndex = 0
-  dataHashCode = -212376212
-  size = 20
-  isKeyFrame = true
-  presentationTimeUs = 0
-sample:
-  trackIndex = 0
-  dataHashCode = -1948569090
-  size = 72
-  isKeyFrame = true
-  presentationTimeUs = 417
-sample:
-  trackIndex = 0
-  dataHashCode = -1316750072
-  size = 84
-  isKeyFrame = true
-  presentationTimeUs = 1917
-sample:
-  trackIndex = 0
-  dataHashCode = 1016428949
-  size = 88
-  isKeyFrame = true
-  presentationTimeUs = 3667
-sample:
-  trackIndex = 0
-  dataHashCode = -1127325245
-  size = 96
-  isKeyFrame = true
-  presentationTimeUs = 5500
-sample:
-  trackIndex = 0
-  dataHashCode = 1148147726
-  size = 92
-  isKeyFrame = true
-  presentationTimeUs = 7500
-sample:
-  trackIndex = 0
-  dataHashCode = -2125685540
-  size = 76
-  isKeyFrame = true
-  presentationTimeUs = 9417
-sample:
-  trackIndex = 0
-  dataHashCode = 473329679
-  size = 24
-  isKeyFrame = true
-  presentationTimeUs = 11000
-sample:
-  trackIndex = 0
-  dataHashCode = 240990900
-  size = 176
-  isKeyFrame = true
-  presentationTimeUs = 11500
-sample:
-  trackIndex = 0
-  dataHashCode = 777637182
-  size = 196
-  isKeyFrame = true
-  presentationTimeUs = 15167
-sample:
-  trackIndex = 0
-  dataHashCode = 1872106264
-  size = 180
-  isKeyFrame = true
-  presentationTimeUs = 19250
-sample:
-  trackIndex = 0
-  dataHashCode = -1520711499
-  size = 140
-  isKeyFrame = true
-  presentationTimeUs = 23000
 sample:
   trackIndex = 0
   dataHashCode = 1580199067
diff --git a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_with_increasing_timestamps_320w_240h.mp4.clipped.dump b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_with_increasing_timestamps_320w_240h.mp4.clipped.dump
index f37630c1cd..ec0af92443 100644
--- a/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_with_increasing_timestamps_320w_240h.mp4.clipped.dump
+++ b/libraries/test_data/src/test/assets/transformerdumps/mp4/sample_with_increasing_timestamps_320w_240h.mp4.clipped.dump
@@ -27,6 +27,12 @@ format 1:
   initializationData:
     data = length 31, hash 4B108214
     data = length 9, hash FBA158BB
+sample:
+  trackIndex = 0
+  dataHashCode = 620415738
+  size = 508
+  isKeyFrame = true
+  presentationTimeUs = 7020
 sample:
   trackIndex = 1
   dataHashCode = 983000500
@@ -219,18 +225,18 @@ sample:
   size = 151
   isKeyFrame = false
   presentationTimeUs = 516666
-sample:
-  trackIndex = 0
-  dataHashCode = 620415738
-  size = 508
-  isKeyFrame = true
-  presentationTimeUs = 7020
 sample:
   trackIndex = 0
   dataHashCode = 33931768
   size = 504
   isKeyFrame = true
   presentationTimeUs = 28354
+sample:
+  trackIndex = 1
+  dataHashCode = 264118578
+  size = 2235
+  isKeyFrame = false
+  presentationTimeUs = 533333
 sample:
   trackIndex = 0
   dataHashCode = 800699278
@@ -370,17 +376,23 @@ sample:
   isKeyFrame = true
   presentationTimeUs = 519020
 sample:
-  trackIndex = 1
-  dataHashCode = 264118578
-  size = 2235
-  isKeyFrame = false
-  presentationTimeUs = 533333
+  trackIndex = 0
+  dataHashCode = 204379389
+  size = 504
+  isKeyFrame = true
+  presentationTimeUs = 540354
 sample:
   trackIndex = 1
   dataHashCode = 64254117
   size = 164
   isKeyFrame = false
   presentationTimeUs = 550000
+sample:
+  trackIndex = 0
+  dataHashCode = 694913274
+  size = 508
+  isKeyFrame = true
+  presentationTimeUs = 561687
 sample:
   trackIndex = 1
   dataHashCode = -1000078879
@@ -556,23 +568,29 @@ sample:
   isKeyFrame = false
   presentationTimeUs = 1033333
 sample:
-  trackIndex = 0
-  dataHashCode = 204379389
-  size = 504
-  isKeyFrame = true
-  presentationTimeUs = 540354
+  trackIndex = 1
+  dataHashCode = 65238903
+  size = 163
+  isKeyFrame = false
+  presentationTimeUs = 1050000
 sample:
-  trackIndex = 0
-  dataHashCode = 694913274
-  size = 508
-  isKeyFrame = true
-  presentationTimeUs = 561687
+  trackIndex = 1
+  dataHashCode = 1720840922
+  size = 2043
+  isKeyFrame = false
+  presentationTimeUs = 1066666
 sample:
   trackIndex = 0
   dataHashCode = 289018778
   size = 513
   isKeyFrame = true
   presentationTimeUs = 583020
+sample:
+  trackIndex = 1
+  dataHashCode = -1006231050
+  size = 178
+  isKeyFrame = false
+  presentationTimeUs = 1083333
 sample:
   trackIndex = 0
   dataHashCode = -693167785
@@ -700,29 +718,35 @@ sample:
   isKeyFrame = true
   presentationTimeUs = 1031020
 sample:
-  trackIndex = 1
-  dataHashCode = 65238903
-  size = 163
-  isKeyFrame = false
-  presentationTimeUs = 1050000
+  trackIndex = 0
+  dataHashCode = -320096195
+  size = 509
+  isKeyFrame = true
+  presentationTimeUs = 1052354
 sample:
-  trackIndex = 1
-  dataHashCode = 1720840922
-  size = 2043
-  isKeyFrame = false
-  presentationTimeUs = 1066666
+  trackIndex = 0
+  dataHashCode = -744850549
+  size = 511
+  isKeyFrame = true
+  presentationTimeUs = 1073687
 sample:
-  trackIndex = 1
-  dataHashCode = -1006231050
-  size = 178
-  isKeyFrame = false
-  presentationTimeUs = 1083333
+  trackIndex = 0
+  dataHashCode = 1457899387
+  size = 505
+  isKeyFrame = true
+  presentationTimeUs = 1095020
 sample:
   trackIndex = 1
   dataHashCode = 1742965952
   size = 2022
   isKeyFrame = false
   presentationTimeUs = 1100000
+sample:
+  trackIndex = 0
+  dataHashCode = 168118808
+  size = 519
+  isKeyFrame = true
+  presentationTimeUs = 1116354
 sample:
   trackIndex = 1
   dataHashCode = -971065365
@@ -861,30 +885,6 @@ sample:
   size = 119
   isKeyFrame = false
   presentationTimeUs = 1483333
-sample:
-  trackIndex = 0
-  dataHashCode = -320096195
-  size = 509
-  isKeyFrame = true
-  presentationTimeUs = 1052354
-sample:
-  trackIndex = 0
-  dataHashCode = -744850549
-  size = 511
-  isKeyFrame = true
-  presentationTimeUs = 1073687
-sample:
-  trackIndex = 0
-  dataHashCode = 1457899387
-  size = 505
-  isKeyFrame = true
-  presentationTimeUs = 1095020
-sample:
-  trackIndex = 0
-  dataHashCode = 168118808
-  size = 519
-  isKeyFrame = true
-  presentationTimeUs = 1116354
 sample:
   trackIndex = 0
   dataHashCode = 896298799
diff --git a/libraries/transformer/src/main/java/androidx/media3/transformer/TransformerInternal.java b/libraries/transformer/src/main/java/androidx/media3/transformer/TransformerInternal.java
index b6906c68fa..189cc082ee 100644
--- a/libraries/transformer/src/main/java/androidx/media3/transformer/TransformerInternal.java
+++ b/libraries/transformer/src/main/java/androidx/media3/transformer/TransformerInternal.java
@@ -17,7 +17,6 @@
 package androidx.media3.transformer;
 
 import static androidx.media3.common.util.Assertions.checkNotNull;
-import static androidx.media3.common.util.Assertions.checkStateNotNull;
 import static androidx.media3.transformer.TransformationException.ERROR_CODE_MUXING_FAILED;
 import static androidx.media3.transformer.Transformer.PROGRESS_STATE_AVAILABLE;
 import static androidx.media3.transformer.Transformer.PROGRESS_STATE_UNAVAILABLE;
@@ -91,7 +90,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
   private static final int MSG_DRAIN_PIPELINES = 4;
   private static final int MSG_END = 5;
 
-  private static final int DRAIN_PIPELINES_DELAY_MS = 10;
+  private static final int DRAIN_PIPELINES_DELAY_MS = 50;
 
   private final Context context;
   private final TransformationRequest transformationRequest;
@@ -223,13 +222,13 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
           startInternal();
           break;
         case MSG_REGISTER_SAMPLE_PIPELINE:
-          samplePipelines.add((SamplePipeline) msg.obj);
+          registerSamplePipelineInternal((SamplePipeline) msg.obj);
           break;
         case MSG_DEQUEUE_INPUT:
           dequeueInputInternal(/* samplePipelineIndex= */ msg.arg1);
           break;
         case MSG_QUEUE_INPUT:
-          queueInputInternal(/* samplePipelineIndex= */ msg.arg1);
+          samplePipelines.get(/* samplePipelineIndex= */ msg.arg1).queueInputBuffer();
           break;
         case MSG_DRAIN_PIPELINES:
           drainPipelinesInternal();
@@ -254,12 +253,19 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
     exoPlayerAssetLoader.start();
   }
 
+  private void registerSamplePipelineInternal(SamplePipeline samplePipeline) {
+    samplePipelines.add(samplePipeline);
+    if (!isDrainingPipelines) {
+      // Make sure pipelines are drained regularly to prevent them from getting stuck.
+      internalHandler.sendEmptyMessageDelayed(MSG_DRAIN_PIPELINES, DRAIN_PIPELINES_DELAY_MS);
+      isDrainingPipelines = true;
+    }
+  }
+
   private void dequeueInputInternal(int samplePipelineIndex) throws TransformationException {
     SamplePipeline samplePipeline = samplePipelines.get(samplePipelineIndex);
-    // The sample pipeline is drained before dequeuing input. It can't be done before queuing
-    // input because, if the pipeline is full, dequeuing input would forever return a null buffer.
-    // Draining the pipeline at regular intervals would be inefficient because a low interval could
-    // result in many no-op operations, and a high interval could slow down data queuing.
+    // The sample pipeline is drained before dequeuing input to maximise the chances of having an
+    // input buffer to dequeue.
     while (samplePipeline.processData()) {}
     pendingInputBuffer = samplePipeline.dequeueInputBuffer();
     dequeueBufferConditionVariable.open();
@@ -269,15 +275,6 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
     }
   }
 
-  private void queueInputInternal(int samplePipelineIndex) throws TransformationException {
-    DecoderInputBuffer pendingInputBuffer = checkStateNotNull(this.pendingInputBuffer);
-    samplePipelines.get(samplePipelineIndex).queueInputBuffer();
-    if (pendingInputBuffer.isEndOfStream() && !isDrainingPipelines) {
-      internalHandler.sendEmptyMessageDelayed(MSG_DRAIN_PIPELINES, DRAIN_PIPELINES_DELAY_MS);
-      isDrainingPipelines = true;
-    }
-  }
-
   private void drainPipelinesInternal() throws TransformationException {
     for (int i = 0; i < samplePipelines.size(); i++) {
       while (samplePipelines.get(i).processData()) {}
@@ -630,7 +627,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
           return null;
         }
         // TODO(b/252537210): Reduce the number of thread hops (for example by adding a queue at the
-        //  start of thesample pipelines). Having 2 thread hops per sample (one for dequeuing and
+        //  start of the sample pipelines). Having 2 thread hops per sample (one for dequeuing and
         //  one for queuing) makes transmuxing slower than it used to be.
         internalHandler
             .obtainMessage(MSG_DEQUEUE_INPUT, samplePipelineIndex, /* unused */ 0)