diff --git a/content/media/webm/WebMReader.cpp b/content/media/webm/WebMReader.cpp
index 2429d0b8f5f..8b2eb1749df 100644
--- a/content/media/webm/WebMReader.cpp
+++ b/content/media/webm/WebMReader.cpp
@@ -372,7 +372,7 @@ nsresult WebMReader::ReadMetadata(MediaInfo* aInfo,
       mHasAudio = true;
       mInfo.mAudio.mHasAudio = true;
       mAudioCodec = nestegg_track_codec_id(mContext, track);
-      mCodecDelay = params.codec_delay;
+      mCodecDelay = params.codec_delay / NS_PER_USEC;
 
       if (mAudioCodec == NESTEGG_CODEC_VORBIS) {
         // Get the Vorbis header data
@@ -439,6 +439,12 @@ nsresult WebMReader::ReadMetadata(MediaInfo* aInfo,
           return NS_ERROR_FAILURE;
         }
 
+        if (mCodecDelay != FramesToUsecs(mOpusParser->mPreSkip, mOpusParser->mRate).value()) {
+          LOG(PR_LOG_DEBUG, ("Values for CodecDelay and PreSkip do not match\n"))
+          Cleanup();
+          return NS_ERROR_FAILURE;
+        }
+
         mInfo.mAudio.mRate = mOpusParser->mRate;
 
         mInfo.mAudio.mChannels = mOpusParser->mChannels;
@@ -718,7 +724,7 @@ bool WebMReader::DecodeAudioPacket(nestegg_packet* aPacket, int64_t aOffset)
         NS_WARNING("Int overflow converting WebM audio duration");
         return false;
       }
-      CheckedInt64 time = startTime - (mCodecDelay / NS_PER_USEC);
+      CheckedInt64 time = startTime - mCodecDelay;
       if (!time.isValid()) {
         NS_WARNING("Int overflow shifting tstamp by codec delay");
         nestegg_free_packet(aPacket);
diff --git a/content/media/webm/WebMReader.h b/content/media/webm/WebMReader.h
index 79554787a9f..278d3e572e9 100644
--- a/content/media/webm/WebMReader.h
+++ b/content/media/webm/WebMReader.h
@@ -215,7 +215,7 @@ private:
   // Number of audio frames we've decoded since decoding began at mAudioStartMs.
   uint64_t mAudioFrames;
 
-  // Number of nanoseconds that must be discarded from the start of the Stream.
+  // Number of microseconds that must be discarded from the start of the Stream.
   uint64_t mCodecDelay;
 
   // Parser state and computed offset-time mappings.  Shared by multiple