Merge branch 'AlexxIT:master' into tuya-new

pkg/aac/adts.go

@@ -8,8 +8,19 @@ import (
 	"github.com/pion/rtp"
 )
 
+const ADTSHeaderSize = 7
+
 func IsADTS(b []byte) bool {
-	return len(b) > 7 && b[0] == 0xFF && b[1]&0xF6 == 0xF0
+	// AAAAAAAA AAAABCCD EEFFFFGH HHIJKLMM MMMMMMMM MMMOOOOO OOOOOOPP (QQQQQQQQ QQQQQQQQ)
+	// A	12	Syncword, all bits must be set to 1.
+	// C	2	Layer, always set to 0.
+	return len(b) >= ADTSHeaderSize && b[0] == 0xFF && b[1]&0b1111_0110 == 0xF0
+}
+
+func HasCRC(b []byte) bool {
+	// AAAAAAAA AAAABCCD EEFFFFGH HHIJKLMM MMMMMMMM MMMOOOOO OOOOOOPP (QQQQQQQQ QQQQQQQQ)
+	// D	1	Protection absence, set to 1 if there is no CRC and 0 if there is CRC.
+	return b[1]&0b1 == 0
 }
 
 func ADTSToCodec(b []byte) *core.Codec {
@@ -58,7 +69,7 @@ func ADTSToCodec(b []byte) *core.Codec {
 func ReadADTSSize(b []byte) uint16 {
 	// AAAAAAAA AAAABCCD EEFFFFGH HHIJKLMM MMMMMMMM MMMOOOOO OOOOOOPP (QQQQQQQQ QQQQQQQQ)
 	_ = b[5] // bounds
-	return uint16(b[3]&0x03)<<(8+3) | uint16(b[4])<<3 | uint16(b[5]>>5)
+	return uint16(b[3]&0b11)<<11 | uint16(b[4])<<3 | uint16(b[5]>>5)
 }
 
 func WriteADTSSize(b []byte, size uint16) {

pkg/aac/producer.go

@@ -2,7 +2,7 @@ package aac
 
 import (
 	"bufio"
-	"encoding/binary"
+	"errors"
 	"io"
 
 	"github.com/AlexxIT/go2rtc/pkg/core"
@@ -17,16 +17,22 @@ type Producer struct {
 func Open(r io.Reader) (*Producer, error) {
 	rd := bufio.NewReader(r)
 
-	b, err := rd.Peek(8)
+	b, err := rd.Peek(ADTSHeaderSize)
 	if err != nil {
 		return nil, err
 	}
 
+	codec := ADTSToCodec(b)
+	if codec == nil {
+		return nil, errors.New("adts: wrong header")
+	}
+	codec.PayloadType = core.PayloadTypeRAW
+
 	medias := []*core.Media{
 		{
 			Kind:      core.KindAudio,
 			Direction: core.DirectionRecvonly,
-			Codecs:    []*core.Codec{ADTSToCodec(b)},
+			Codecs:    []*core.Codec{codec},
 		},
 	}
 	return &Producer{
@@ -42,14 +48,25 @@ func Open(r io.Reader) (*Producer, error) {
 
 func (c *Producer) Start() error {
 	for {
-		b, err := c.rd.Peek(6)
+		// read ADTS header
-		if err != nil {
+		adts := make([]byte, ADTSHeaderSize)
+		if _, err := io.ReadFull(c.rd, adts); err != nil {
 			return err
 		}
 
-		auSize := ReadADTSSize(b)
+		auSize := ReadADTSSize(adts) - ADTSHeaderSize
-		payload := make([]byte, 2+2+auSize)
+
-		if _, err = io.ReadFull(c.rd, payload[4:]); err != nil {
+		if HasCRC(adts) {
+			// skip CRC after header
+			if _, err := c.rd.Discard(2); err != nil {
+				return err
+			}
+			auSize -= 2
+		}
+
+		// read AAC payload after header
+		payload := make([]byte, auSize)
+		if _, err := io.ReadFull(c.rd, payload); err != nil {
 			return err
 		}
 
@@ -59,9 +76,6 @@ func (c *Producer) Start() error {
 			continue
 		}
 
-		payload[1] = 16 // header size in bits
-		binary.BigEndian.PutUint16(payload[2:], auSize<<3)
-
 		pkt := &rtp.Packet{
 			Header:  rtp.Header{Timestamp: core.Now90000()},
 			Payload: payload,

pkg/aac/rtp.go

@@ -8,7 +8,6 @@ import (
 )
 
 const RTPPacketVersionAAC = 0
-const ADTSHeaderSize = 7
 
 func RTPDepay(handler core.HandlerFunc) core.HandlerFunc {
 	var timestamp uint32
@@ -65,7 +64,8 @@ func RTPDepay(handler core.HandlerFunc) core.HandlerFunc {
 }
 
 func RTPPay(handler core.HandlerFunc) core.HandlerFunc {
-	sequencer := rtp.NewRandomSequencer()
+	var seq uint16
+	var ts uint32
 
 	return func(packet *rtp.Packet) {
 		if packet.Version != RTPPacketVersionAAC {
@@ -85,12 +85,15 @@ func RTPPay(handler core.HandlerFunc) core.HandlerFunc {
 			Header: rtp.Header{
 				Version:        2,
 				Marker:         true,
-				SequenceNumber: sequencer.NextSequenceNumber(),
+				SequenceNumber: seq,
-				Timestamp:      packet.Timestamp,
+				Timestamp:      ts,
 			},
 			Payload: payload,
 		}
 		handler(&clone)
+
+		seq++
+		ts += AUTime
 	}
 }
 

pkg/flv/flv_test.go

@@ -0,0 +1,21 @@
+package flv
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestTimeToRTP(t *testing.T) {
+	// Reolink camera has 20 FPS
+	// Video timestamp increases by 50ms, SampleRate 90000, RTP timestamp increases by 4500
+	// Audio timestamp increases by 64ms, SampleRate 16000, RTP timestamp increases by 1024
+	frameN := 1
+	for i := 0; i < 32; i++ {
+		// 1000ms/(90000/4500) = 50ms
+		require.Equal(t, uint32(frameN*4500), TimeToRTP(uint32(frameN*50), 90000))
+		// 1000ms/(16000/1024) = 64ms
+		require.Equal(t, uint32(frameN*1024), TimeToRTP(uint32(frameN*64), 16000))
+		frameN *= 2
+	}
+}

pkg/flv/producer.go

@@ -299,8 +299,12 @@ func (c *Producer) readPacket() (*rtp.Packet, error) {
 	return pkt, nil
 }
 
-func TimeToRTP(timeMS uint32, clockRate uint32) uint32 {
+// TimeToRTP convert time in milliseconds to RTP time
-	return timeMS * clockRate / 1000
+func TimeToRTP(timeMS, clockRate uint32) uint32 {
+	// for clockRates 90000, 16000, 8000, etc. - we can use:
+	//     return timeMS * (clockRate / 1000)
+	// but for clockRates 44100, 22050, 11025 - we should use:
+	return uint32(uint64(timeMS) * uint64(clockRate) / 1000)
 }
 
 func isExHeader(data []byte) bool {