hls: prefer Opus tracks to MPEG-4 tracks (#2158)

internal/core/hls_muxer.go

@@ -456,20 +456,16 @@ func (m *hlsMuxer) createVideoTrack(stream *stream.Stream) (*media.Media, *gohls
 }
 
 func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohlslib.Track) {
-	var audioFormatMPEG4AudioGeneric *formats.MPEG4AudioGeneric
-	audioMedia := stream.Medias().FindFormat(&audioFormatMPEG4AudioGeneric)
+	var audioFormatOpus *formats.Opus
+	audioMedia := stream.Medias().FindFormat(&audioFormatOpus)
 
 	if audioMedia != nil {
 		audioStartPTSFilled := false
 		var audioStartPTS time.Duration
 
-		stream.AddReader(m, audioMedia, audioFormatMPEG4AudioGeneric, func(unit formatprocessor.Unit) {
+		stream.AddReader(m, audioMedia, audioFormatOpus, func(unit formatprocessor.Unit) {
 			m.ringBuffer.Push(func() error {
-				tunit := unit.(*formatprocessor.UnitMPEG4AudioGeneric)
-
-				if tunit.AUs == nil {
-					return nil
-				}
+				tunit := unit.(*formatprocessor.UnitOpus)
 
 				if !audioStartPTSFilled {
 					audioStartPTSFilled = true
@@ -477,10 +473,10 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 				}
 
 				pts := tunit.PTS - audioStartPTS
-				err := m.muxer.WriteMPEG4Audio(
+				err := m.muxer.WriteOpus(
 					tunit.NTP,
 					pts,
-					tunit.AUs)
+					tunit.Packets)
 				if err != nil {
 					return fmt.Errorf("muxer error: %v", err)
 				}
@@ -490,27 +486,29 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 		})
 
 		return audioMedia, &gohlslib.Track{
-			Codec: &codecs.MPEG4Audio{
-				Config: *audioFormatMPEG4AudioGeneric.Config,
+			Codec: &codecs.Opus{
+				ChannelCount: func() int {
+					if audioFormatOpus.IsStereo {
+						return 2
+					}
+					return 1
+				}(),
 			},
 		}
 	}
 
-	var audioFormatMPEG4AudioLATM *formats.MPEG4AudioLATM
-	audioMedia = stream.Medias().FindFormat(&audioFormatMPEG4AudioLATM)
+	var audioFormatMPEG4AudioGeneric *formats.MPEG4AudioGeneric
+	audioMedia = stream.Medias().FindFormat(&audioFormatMPEG4AudioGeneric)
 
-	if audioMedia != nil &&
-		audioFormatMPEG4AudioLATM.Config != nil &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs) == 1 &&
-		len(audioFormatMPEG4AudioLATM.Config.Programs[0].Layers) == 1 {
+	if audioMedia != nil {
 		audioStartPTSFilled := false
 		var audioStartPTS time.Duration
 
-		stream.AddReader(m, audioMedia, audioFormatMPEG4AudioLATM, func(unit formatprocessor.Unit) {
+		stream.AddReader(m, audioMedia, audioFormatMPEG4AudioGeneric, func(unit formatprocessor.Unit) {
 			m.ringBuffer.Push(func() error {
-				tunit := unit.(*formatprocessor.UnitMPEG4AudioLATM)
+				tunit := unit.(*formatprocessor.UnitMPEG4AudioGeneric)
 
-				if tunit.AU == nil {
+				if tunit.AUs == nil {
 					return nil
 				}
 
@@ -523,7 +521,7 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 				err := m.muxer.WriteMPEG4Audio(
 					tunit.NTP,
 					pts,
-					[][]byte{tunit.AU})
+					tunit.AUs)
 				if err != nil {
 					return fmt.Errorf("muxer error: %v", err)
 				}
@@ -534,21 +532,28 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 
 		return audioMedia, &gohlslib.Track{
 			Codec: &codecs.MPEG4Audio{
-				Config: *audioFormatMPEG4AudioLATM.Config.Programs[0].Layers[0].AudioSpecificConfig,
+				Config: *audioFormatMPEG4AudioGeneric.Config,
 			},
 		}
 	}
 
-	var audioFormatOpus *formats.Opus
-	audioMedia = stream.Medias().FindFormat(&audioFormatOpus)
+	var audioFormatMPEG4AudioLATM *formats.MPEG4AudioLATM
+	audioMedia = stream.Medias().FindFormat(&audioFormatMPEG4AudioLATM)
 
-	if audioMedia != nil {
+	if audioMedia != nil &&
+		audioFormatMPEG4AudioLATM.Config != nil &&
+		len(audioFormatMPEG4AudioLATM.Config.Programs) == 1 &&
+		len(audioFormatMPEG4AudioLATM.Config.Programs[0].Layers) == 1 {
 		audioStartPTSFilled := false
 		var audioStartPTS time.Duration
 
-		stream.AddReader(m, audioMedia, audioFormatOpus, func(unit formatprocessor.Unit) {
+		stream.AddReader(m, audioMedia, audioFormatMPEG4AudioLATM, func(unit formatprocessor.Unit) {
 			m.ringBuffer.Push(func() error {
-				tunit := unit.(*formatprocessor.UnitOpus)
+				tunit := unit.(*formatprocessor.UnitMPEG4AudioLATM)
+
+				if tunit.AU == nil {
+					return nil
+				}
 
 				if !audioStartPTSFilled {
 					audioStartPTSFilled = true
@@ -556,10 +561,10 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 				}
 
 				pts := tunit.PTS - audioStartPTS
-				err := m.muxer.WriteOpus(
+				err := m.muxer.WriteMPEG4Audio(
 					tunit.NTP,
 					pts,
-					tunit.Packets)
+					[][]byte{tunit.AU})
 				if err != nil {
 					return fmt.Errorf("muxer error: %v", err)
 				}
@@ -569,13 +574,8 @@ func (m *hlsMuxer) createAudioTrack(stream *stream.Stream) (*media.Media, *gohls
 		})
 
 		return audioMedia, &gohlslib.Track{
-			Codec: &codecs.Opus{
-				ChannelCount: func() int {
-					if audioFormatOpus.IsStereo {
-						return 2
-					}
-					return 1
-				}(),
+			Codec: &codecs.MPEG4Audio{
+				Config: *audioFormatMPEG4AudioLATM.Config.Programs[0].Layers[0].AudioSpecificConfig,
 			},
 		}
 	}