diff --git a/packager/media/formats/mp4/box_definitions.cc b/packager/media/formats/mp4/box_definitions.cc
index f28e287244..87a236d2b1 100644
--- a/packager/media/formats/mp4/box_definitions.cc
+++ b/packager/media/formats/mp4/box_definitions.cc
@@ -1895,7 +1895,10 @@ size_t WebVTTSourceLabelBox::ComputeSizeInternal() {
   return HeaderSize() + source_label.size();
 }
 
-TextSampleEntry::TextSampleEntry() : format(FOURCC_NULL) {}
+// data_reference_index is 1-based and "dref" box is mandatory so it is
+// always present.
+TextSampleEntry::TextSampleEntry()
+    : format(FOURCC_NULL), data_reference_index(1u) {}
 TextSampleEntry::~TextSampleEntry() {}
 
 FourCC TextSampleEntry::BoxType() const {
diff --git a/packager/media/formats/mp4/box_definitions.h b/packager/media/formats/mp4/box_definitions.h
index 2c208594b0..a264eee185 100644
--- a/packager/media/formats/mp4/box_definitions.h
+++ b/packager/media/formats/mp4/box_definitions.h
@@ -377,6 +377,7 @@ struct TextSampleEntry : Box {
   // set to 'wvtt' to write WVTTSampleEntry; On read, it is recovered from box
   // header.
   FourCC format;
+
   uint16_t data_reference_index;
 
   // Sub boxes for wvtt text sample entry.