install go2rtc on bob

2026-04-04 19:36:14 +02:00
parent f0b56e63d1
commit ccf88187b8
537 changed files with 69213 additions and 0 deletions
@@ -0,0 +1,5 @@
+## Useful links
+
+- [RFC 3550: RTP: A Transport Protocol for Real-Time Applications](https://datatracker.ietf.org/doc/html/rfc3550)
+- [RFC 6716: Definition of the Opus Audio Codec](https://datatracker.ietf.org/doc/html/rfc6716)
+- [RFC 7587: RTP Payload Format for the Opus Speech and Audio Codec](https://datatracker.ietf.org/doc/html/rfc7587)
@@ -0,0 +1,96 @@
+package opus
+
+import (
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/pion/rtp"
+)
+
+// Some info about this magic:
+// - Apple has no respect for RFC 7587 standard and using RFC 3550 for RTP timestamps
+// - Apple can request packets with 20ms duration over LAN connection and 60ms over LTE
+// - FFmpeg produce packets with 20ms duration by default and only one frame per packet
+// - FFmpeg should use "-min_comp 0" option, so every packet will be same duration
+// - Apple doesn't care about real sample rate of track
+// - Apple only cares about proper timestamp based on REQUESTED sample rate
+
+// RepackToHAP - convert standart RTP packet with OPUS to HAP packet
+// We expect that:
+// - incoming packet will be 20ms duration and only one frame per packet
+// - outgouing packet will be 20ms or 60ms duration
+// - incoming sample rate will be any (but not very big if we needs 60ms packets for output)
+// - outgouing sample rate will be 16000
+// https://github.com/AlexxIT/go2rtc/issues/667
+func RepackToHAP(rtpTime byte, handler core.HandlerFunc) core.HandlerFunc {
+	switch rtpTime {
+	case 20:
+		return repackToHAP20(handler)
+	case 60:
+		return repackToHAP60(handler)
+	}
+	return handler
+}
+
+// we using only one sample rate in the pkg/hap/camera/accessory.go
+const (
+	timestamp20 = 16000 * 0.020
+	timestamp60 = 16000 * 0.060
+)
+
+// repackToHAP20 - just fix RTP timestamp from RFC 7587 to RFC 3550
+func repackToHAP20(handler core.HandlerFunc) core.HandlerFunc {
+	var timestamp uint32
+
+	return func(pkt *rtp.Packet) {
+		timestamp += timestamp20
+
+		clone := *pkt
+		clone.Timestamp = timestamp
+		handler(&clone)
+	}
+}
+
+// repackToHAP60 - collect 20ms frames to single 60ms packet
+// thanks to @civita idea https://github.com/AlexxIT/go2rtc/pull/843
+func repackToHAP60(handler core.HandlerFunc) core.HandlerFunc {
+	var sequence uint16
+	var timestamp uint32
+
+	var framesCount byte
+	var framesSize []byte
+	var framesData []byte
+
+	return func(pkt *rtp.Packet) {
+		framesData = append(framesData, pkt.Payload[1:]...)
+
+		if framesCount++; framesCount < 3 {
+			if frameSize := len(pkt.Payload) - 1; frameSize >= 252 {
+				b0 := 252 + byte(frameSize)&0b11
+				framesSize = append(framesSize, b0, byte(frameSize/4)-b0)
+			} else {
+				framesSize = append(framesSize, byte(frameSize))
+			}
+			return
+		}
+
+		toc := pkt.Payload[0]
+
+		payload := make([]byte, 2, 2+len(framesSize)+len(framesData))
+		payload[0] = toc | 0b11  // code 3 (multiple frames per packet)
+		payload[1] = 0b1000_0011 // VBR, no padding, 3 frames
+		payload = append(payload, framesSize...)
+		payload = append(payload, framesData...)
+
+		sequence++
+		timestamp += timestamp60
+
+		clone := *pkt
+		clone.Payload = payload
+		clone.SequenceNumber = sequence
+		clone.Timestamp = timestamp
+		handler(&clone)
+
+		framesCount = 0
+		framesSize = framesSize[:0]
+		framesData = framesData[:0]
+	}
+}
@@ -0,0 +1,118 @@
+package opus
+
+import (
+	"time"
+)
+
+type Header struct {
+	Mode       string
+	SampleRate uint16
+	FrameSize  time.Duration
+	Channels   byte
+	Frames     byte
+}
+
+func UnmarshalHeader(b []byte) *Header {
+	// https://datatracker.ietf.org/doc/html/rfc6716#section-3.1
+	b0 := b[0]
+	config := b0 >> 3
+	return &Header{
+		Mode:       parseMode(config),
+		SampleRate: parseSampleRate(config),
+		FrameSize:  parseFrameSize(config),
+		Channels:   parseChannels(b0 >> 2 & 0b1),
+		Frames:     parseFrames(b0 & 0b11),
+	}
+}
+
+func parseMode(config byte) string {
+	if config <= 11 {
+		return "silk"
+	}
+	if config <= 15 {
+		return "hybrid"
+	}
+	return "celt"
+}
+
+func parseSampleRate(config byte) uint16 {
+	switch config {
+	case 0, 1, 2, 3, 16, 17, 18, 19:
+		return 8000 // NB (narrowband)
+	case 4, 5, 6, 7:
+		return 12000 // MB (medium-band)
+	case 8, 9, 10, 11, 20, 21, 22, 23:
+		return 16000 // WB (wideband)
+	case 12, 13, 24, 25, 26, 27:
+		return 24000 // SWB (super-wideband)
+	case 14, 15, 28, 29, 30, 31:
+		return 48000 // FB (fullband)
+	}
+	return 0
+}
+
+func parseFrameSize(config byte) time.Duration {
+	switch config {
+	case 0, 4, 8, 12, 14, 18, 22, 26, 30:
+		return 10_000_000
+	case 1, 5, 9, 13, 15, 19, 23, 27, 31:
+		return 20_000_000
+	case 2, 6, 10:
+		return 40_000_000
+	case 3, 7, 11:
+		return 60_000_000
+	case 16, 20, 24, 28:
+		return 2_500_000
+	case 17, 21, 25, 29:
+		return 5_000_000
+	}
+	return 0
+}
+
+func parseChannels(s byte) byte {
+	if s == 1 {
+		return 2
+	}
+	return 1
+}
+
+func parseFrames(c byte) byte {
+	switch c {
+	case 0:
+		return 1
+	case 1, 2:
+		return 2
+	}
+	return 0xFF
+}
+
+func JoinFrames(b1, b2 []byte) []byte {
+	// can't join
+	if b1[0]&0b11 != 0 || b2[0]&0b11 != 0 {
+		return append(b1, b2...)
+	}
+
+	size1, size2 := len(b1)-1, len(b2)-1
+
+	// join same sizes
+	if size1 == size2 {
+		b := make([]byte, 1+size1+size2)
+		copy(b, b1)
+		copy(b[1+size1:], b2[1:])
+		b[0] |= 0b01
+		return b
+	}
+
+	b := make([]byte, 1, 3+size1+size2)
+	b[0] = b1[0] | 0b10
+	if size1 >= 252 {
+		b0 := 252 + byte(size1)&0b11
+		b = append(b, b0, byte(size1/4)-b0)
+	} else {
+		b = append(b, byte(size1))
+	}
+
+	b = append(b, b1[1:]...)
+	b = append(b, b2[1:]...)
+	return b
+}