support publishing, reading and proxying MPEG-2 audio (MP3) tracks with RTMP (#1102) (#1736)

internal/formatprocessor/generic_test.go

@@ -11,7 +11,7 @@ import (
 func TestGenericRemovePadding(t *testing.T) {
 	forma := &formats.Generic{
 		PayloadTyp: 96,
-		RTPMap:     "private/90000",
+		RTPMa:      "private/90000",
 	}
 	forma.Init()
 

internal/formatprocessor/h264.go

@@ -104,7 +104,12 @@ func newH264(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtph264.Encoder{
+			PayloadMaxSize:    udpMaxPayloadSize - 12,
+			PayloadType:       forma.PayloadTyp,
+			PacketizationMode: forma.PacketizationMode,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -280,6 +285,7 @@ func (t *formatProcessorH264) Process(unit Unit, hasNonRTSPReaders bool) error {
 		tunit.AU = t.remuxAccessUnit(tunit.AU)
 	}
 
+	// encode into RTP
 	if len(tunit.AU) != 0 {
 		pkts, err := t.encoder.Encode(tunit.AU, tunit.PTS)
 		if err != nil {

internal/formatprocessor/h265.go

@@ -111,7 +111,11 @@ func newH265(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtph265.Encoder{
+			PayloadMaxSize: t.udpMaxPayloadSize - 12,
+			PayloadType:    forma.PayloadTyp,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -303,6 +307,7 @@ func (t *formatProcessorH265) Process(unit Unit, hasNonRTSPReaders bool) error {
 		tunit.AU = t.remuxAccessUnit(tunit.AU)
 	}
 
+	// encode into RTP
 	if len(tunit.AU) != 0 {
 		pkts, err := t.encoder.Encode(tunit.AU, tunit.PTS)
 		if err != nil {

internal/formatprocessor/mpeg2audio.go

@@ -0,0 +1,99 @@
+package formatprocessor
+
+import (
+	"fmt"
+	"time"
+
+	"github.com/bluenviron/gortsplib/v3/pkg/formats"
+	"github.com/bluenviron/gortsplib/v3/pkg/formats/rtpmpeg2audio"
+	"github.com/pion/rtp"
+)
+
+// UnitMPEG2Audio is a MPEG2-audio data unit.
+type UnitMPEG2Audio struct {
+	RTPPackets []*rtp.Packet
+	NTP        time.Time
+	PTS        time.Duration
+	Frames     [][]byte
+}
+
+// GetRTPPackets implements Unit.
+func (d *UnitMPEG2Audio) GetRTPPackets() []*rtp.Packet {
+	return d.RTPPackets
+}
+
+// GetNTP implements Unit.
+func (d *UnitMPEG2Audio) GetNTP() time.Time {
+	return d.NTP
+}
+
+type formatProcessorMPEG2Audio struct {
+	udpMaxPayloadSize int
+	format            *formats.MPEG2Audio
+	encoder           *rtpmpeg2audio.Encoder
+	decoder           *rtpmpeg2audio.Decoder
+}
+
+func newMPEG2Audio(
+	udpMaxPayloadSize int,
+	forma *formats.MPEG2Audio,
+	allocateEncoder bool,
+) (*formatProcessorMPEG2Audio, error) {
+	t := &formatProcessorMPEG2Audio{
+		udpMaxPayloadSize: udpMaxPayloadSize,
+		format:            forma,
+	}
+
+	if allocateEncoder {
+		t.encoder = &rtpmpeg2audio.Encoder{
+			PayloadMaxSize: t.udpMaxPayloadSize - 12,
+		}
+		t.encoder.Init()
+	}
+
+	return t, nil
+}
+
+func (t *formatProcessorMPEG2Audio) Process(unit Unit, hasNonRTSPReaders bool) error { //nolint:dupl
+	tunit := unit.(*UnitMPEG2Audio)
+
+	if tunit.RTPPackets != nil {
+		pkt := tunit.RTPPackets[0]
+
+		// remove padding
+		pkt.Header.Padding = false
+		pkt.PaddingSize = 0
+
+		if pkt.MarshalSize() > t.udpMaxPayloadSize {
+			return fmt.Errorf("payload size (%d) is greater than maximum allowed (%d)",
+				pkt.MarshalSize(), t.udpMaxPayloadSize)
+		}
+
+		// decode from RTP
+		if hasNonRTSPReaders {
+			if t.decoder == nil {
+				t.decoder = t.format.CreateDecoder()
+			}
+
+			frames, pts, err := t.decoder.Decode(pkt)
+			if err != nil {
+				return err
+			}
+
+			tunit.Frames = frames
+			tunit.PTS = pts
+		}
+
+		// route packet as is
+		return nil
+	}
+
+	// encode into RTP
+	pkts, err := t.encoder.Encode(tunit.Frames, tunit.PTS)
+	if err != nil {
+		return err
+	}
+	tunit.RTPPackets = pkts
+
+	return nil
+}

internal/formatprocessor/mpeg4audio.go

@@ -45,7 +45,15 @@ func newMPEG4Audio(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtpmpeg4audio.Encoder{
+			PayloadMaxSize:   t.udpMaxPayloadSize - 12,
+			PayloadType:      forma.PayloadTyp,
+			SampleRate:       forma.Config.SampleRate,
+			SizeLength:       forma.SizeLength,
+			IndexLength:      forma.IndexLength,
+			IndexDeltaLength: forma.IndexDeltaLength,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -88,11 +96,12 @@ func (t *formatProcessorMPEG4Audio) Process(unit Unit, hasNonRTSPReaders bool) e
 		return nil
 	}
 
+	// encode into RTP
 	pkts, err := t.encoder.Encode(tunit.AUs, tunit.PTS)
 	if err != nil {
 		return err
 	}
-
 	tunit.RTPPackets = pkts
+
 	return nil
 }

internal/formatprocessor/opus.go

@@ -45,7 +45,12 @@ func newOpus(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtpsimpleaudio.Encoder{
+			PayloadMaxSize: t.udpMaxPayloadSize - 12,
+			PayloadType:    forma.PayloadTyp,
+			SampleRate:     48000,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -85,11 +90,12 @@ func (t *formatProcessorOpus) Process(unit Unit, hasNonRTSPReaders bool) error {
 		return nil
 	}
 
+	// encode into RTP
 	pkt, err := t.encoder.Encode(tunit.Frame, tunit.PTS)
 	if err != nil {
 		return err
 	}
-
 	tunit.RTPPackets = []*rtp.Packet{pkt}
+
 	return nil
 }

internal/formatprocessor/processor.go

@@ -30,6 +30,9 @@ func New(
 	case *formats.VP9:
 		return newVP9(udpMaxPayloadSize, forma, generateRTPPackets)
 
+	case *formats.MPEG2Audio:
+		return newMPEG2Audio(udpMaxPayloadSize, forma, generateRTPPackets)
+
 	case *formats.MPEG4Audio:
 		return newMPEG4Audio(udpMaxPayloadSize, forma, generateRTPPackets)
 

internal/formatprocessor/vp8.go

@@ -45,7 +45,11 @@ func newVP8(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtpvp8.Encoder{
+			PayloadMaxSize: t.udpMaxPayloadSize - 12,
+			PayloadType:    forma.PayloadTyp,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -88,11 +92,12 @@ func (t *formatProcessorVP8) Process(unit Unit, hasNonRTSPReaders bool) error {
 		return nil
 	}
 
+	// encode into RTP
 	pkts, err := t.encoder.Encode(tunit.Frame, tunit.PTS)
 	if err != nil {
 		return err
 	}
-
 	tunit.RTPPackets = pkts
+
 	return nil
 }

internal/formatprocessor/vp9.go

@@ -45,7 +45,11 @@ func newVP9(
 	}
 
 	if allocateEncoder {
-		t.encoder = forma.CreateEncoder()
+		t.encoder = &rtpvp9.Encoder{
+			PayloadMaxSize: t.udpMaxPayloadSize - 12,
+			PayloadType:    forma.PayloadTyp,
+		}
+		t.encoder.Init()
 	}
 
 	return t, nil
@@ -88,11 +92,12 @@ func (t *formatProcessorVP9) Process(unit Unit, hasNonRTSPReaders bool) error {
 		return nil
 	}
 
+	// encode into RTP
 	pkts, err := t.encoder.Encode(tunit.Frame, tunit.PTS)
 	if err != nil {
 		return err
 	}
-
 	tunit.RTPPackets = pkts
+
 	return nil
 }