install go2rtc on bob

This commit is contained in:
2026-04-04 19:36:14 +02:00
parent f0b56e63d1
commit ccf88187b8
537 changed files with 69213 additions and 0 deletions
+36
View File
@@ -0,0 +1,36 @@
## Fragmented MP4
```
ffmpeg -i "rtsp://..." -movflags +frag_keyframe+separate_moof+default_base_moof+empty_moov -frag_duration 1 -c copy -t 5 sample.mp4
```
- movflags frag_keyframe
Start a new fragment at each video keyframe.
- frag_duration duration
Create fragments that are duration microseconds long.
- movflags separate_moof
Write a separate moof (movie fragment) atom for each track.
- movflags default_base_moof
Similarly to the omit_tfhd_offset, this flag avoids writing the absolute base_data_offset field in tfhd atoms, but does so by using the new default-base-is-moof flag instead.
https://ffmpeg.org/ffmpeg-formats.html#Options-13
## HEVC
| Browser | avc1 | hvc1 | hev1 |
|-------------|------|------|------|
| Mac Chrome | + | - | + |
| Mac Safari | + | + | - |
| iOS 15? | + | + | - |
| Mac Firefox | + | - | - |
| iOS 12 | + | - | - |
| Android 13 | + | - | - |
## Useful links
- https://stackoverflow.com/questions/63468587/what-hevc-codec-tag-to-use-with-fmp4-hvc1-or-hev1
- https://stackoverflow.com/questions/32152090/encode-h265-to-hvc1-codec
- https://jellyfin.org/docs/general/clients/codec-support.html
- https://github.com/StaZhu/enable-chromium-hevc-hardware-decoding
- https://developer.mozilla.org/ru/docs/Web/Media/Formats/codecs_parameter
- https://gstreamer-devel.narkive.com/rhkUolp2/rtp-dts-pts-result-in-varying-mp4-frame-durations
+189
View File
@@ -0,0 +1,189 @@
package mp4
import (
"errors"
"io"
"sync"
"github.com/AlexxIT/go2rtc/pkg/aac"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264"
"github.com/AlexxIT/go2rtc/pkg/h265"
"github.com/AlexxIT/go2rtc/pkg/pcm"
"github.com/pion/rtp"
)
type Consumer struct {
core.Connection
wr *core.WriteBuffer
muxer *Muxer
mu sync.Mutex
start bool
Rotate int `json:"-"`
ScaleX int `json:"-"`
ScaleY int `json:"-"`
}
func NewConsumer(medias []*core.Media) *Consumer {
if medias == nil {
// default local medias
medias = []*core.Media{
{
Kind: core.KindVideo,
Direction: core.DirectionSendonly,
Codecs: []*core.Codec{
{Name: core.CodecH264},
{Name: core.CodecH265},
},
},
{
Kind: core.KindAudio,
Direction: core.DirectionSendonly,
Codecs: []*core.Codec{
{Name: core.CodecAAC},
},
},
}
}
wr := core.NewWriteBuffer(nil)
return &Consumer{
Connection: core.Connection{
ID: core.NewID(),
FormatName: "mp4",
Medias: medias,
Transport: wr,
},
muxer: &Muxer{},
wr: wr,
}
}
func (c *Consumer) AddTrack(media *core.Media, _ *core.Codec, track *core.Receiver) error {
trackID := byte(len(c.Senders))
codec := track.Codec.Clone()
handler := core.NewSender(media, codec)
switch track.Codec.Name {
case core.CodecH264:
handler.Handler = func(packet *rtp.Packet) {
if !c.start {
if !h264.IsKeyframe(packet.Payload) {
return
}
c.start = true
}
// important to use Mutex because right fragment order
c.mu.Lock()
b := c.muxer.GetPayload(trackID, packet)
if n, err := c.wr.Write(b); err == nil {
c.Send += n
}
c.mu.Unlock()
}
if track.Codec.IsRTP() {
handler.Handler = h264.RTPDepay(track.Codec, handler.Handler)
} else {
handler.Handler = h264.RepairAVCC(track.Codec, handler.Handler)
}
case core.CodecH265:
handler.Handler = func(packet *rtp.Packet) {
if !c.start {
if !h265.IsKeyframe(packet.Payload) {
return
}
c.start = true
}
// important to use Mutex because right fragment order
c.mu.Lock()
b := c.muxer.GetPayload(trackID, packet)
if n, err := c.wr.Write(b); err == nil {
c.Send += n
}
c.mu.Unlock()
}
if track.Codec.IsRTP() {
handler.Handler = h265.RTPDepay(track.Codec, handler.Handler)
} else {
handler.Handler = h265.RepairAVCC(track.Codec, handler.Handler)
}
default:
handler.Handler = func(packet *rtp.Packet) {
if !c.start {
return
}
// important to use Mutex because right fragment order
c.mu.Lock()
b := c.muxer.GetPayload(trackID, packet)
if n, err := c.wr.Write(b); err == nil {
c.Send += n
}
c.mu.Unlock()
}
switch track.Codec.Name {
case core.CodecAAC:
if track.Codec.IsRTP() {
handler.Handler = aac.RTPDepay(handler.Handler)
}
case core.CodecOpus, core.CodecMP3: // no changes
case core.CodecPCMA, core.CodecPCMU, core.CodecPCM, core.CodecPCML:
codec.Name = core.CodecFLAC
if codec.Channels == 2 {
// hacky way for support two channels audio
codec.Channels = 1
codec.ClockRate *= 2
}
handler.Handler = pcm.FLACEncoder(track.Codec.Name, codec.ClockRate, handler.Handler)
default:
handler.Handler = nil
}
}
if handler.Handler == nil {
s := "mp4: unsupported codec: " + track.Codec.String()
println(s)
return errors.New(s)
}
c.muxer.AddTrack(codec)
handler.HandleRTP(track)
c.Senders = append(c.Senders, handler)
return nil
}
func (c *Consumer) WriteTo(wr io.Writer) (int64, error) {
if len(c.Senders) == 1 && c.Senders[0].Codec.IsAudio() {
c.start = true
}
init, err := c.muxer.GetInit()
if err != nil {
return 0, err
}
if c.Rotate != 0 {
PatchVideoRotate(init, c.Rotate)
}
if c.ScaleX != 0 && c.ScaleY != 0 {
PatchVideoScale(init, c.ScaleX, c.ScaleY)
}
if _, err = wr.Write(init); err != nil {
return 0, err
}
return c.wr.WriteTo(wr)
}
+116
View File
@@ -0,0 +1,116 @@
package mp4
import (
"github.com/AlexxIT/go2rtc/pkg/aac"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264"
"github.com/AlexxIT/go2rtc/pkg/iso"
"github.com/pion/rtp"
)
type Demuxer struct {
codecs map[uint32]*core.Codec
timeScales map[uint32]float32
}
func (d *Demuxer) Probe(init []byte) (medias []*core.Media) {
var trackID, timeScale uint32
if d.codecs == nil {
d.codecs = make(map[uint32]*core.Codec)
d.timeScales = make(map[uint32]float32)
}
atoms, _ := iso.DecodeAtoms(init)
for _, atom := range atoms {
var codec *core.Codec
switch atom := atom.(type) {
case *iso.AtomTkhd:
trackID = atom.TrackID
case *iso.AtomMdhd:
timeScale = atom.TimeScale
case *iso.AtomVideo:
switch atom.Name {
case "avc1":
codec = h264.ConfigToCodec(atom.Config)
}
case *iso.AtomAudio:
switch atom.Name {
case "mp4a":
codec = aac.ConfigToCodec(atom.Config)
}
}
if codec != nil {
d.codecs[trackID] = codec
d.timeScales[trackID] = float32(codec.ClockRate) / float32(timeScale)
medias = append(medias, &core.Media{
Kind: codec.Kind(),
Direction: core.DirectionRecvonly,
Codecs: []*core.Codec{codec},
})
}
}
return
}
func (d *Demuxer) GetTrackID(codec *core.Codec) uint32 {
for trackID, c := range d.codecs {
if c == codec {
return trackID
}
}
return 0
}
func (d *Demuxer) Demux(data2 []byte) (trackID uint32, packets []*core.Packet) {
atoms, err := iso.DecodeAtoms(data2)
if err != nil {
return 0, nil
}
var ts uint32
var trun *iso.AtomTrun
var data []byte
for _, atom := range atoms {
switch atom := atom.(type) {
case *iso.AtomTfhd:
trackID = atom.TrackID
case *iso.AtomTfdt:
ts = uint32(atom.DecodeTime)
case *iso.AtomTrun:
trun = atom
case *iso.AtomMdat:
data = atom.Data
}
}
timeScale := d.timeScales[trackID]
if timeScale == 0 {
return 0, nil
}
n := len(trun.SamplesDuration)
packets = make([]*core.Packet, n)
for i := 0; i < n; i++ {
duration := trun.SamplesDuration[i]
size := trun.SamplesSize[i]
// can be SPS, PPS and IFrame in one packet
timestamp := uint32(float32(ts) * timeScale)
packets[i] = &rtp.Packet{
Header: rtp.Header{Timestamp: timestamp},
Payload: data[:size],
}
data = data[size:]
ts += duration
}
return
}
+166
View File
@@ -0,0 +1,166 @@
package mp4
import (
"bytes"
"encoding/binary"
"strings"
"github.com/AlexxIT/go2rtc/pkg/core"
)
// ParseQuery - like usual parse, but with mp4 param handler
func ParseQuery(query map[string][]string) []*core.Media {
if v := query["mp4"]; len(v) != 0 {
medias := []*core.Media{
{
Kind: core.KindVideo,
Direction: core.DirectionSendonly,
Codecs: []*core.Codec{
{Name: core.CodecH264},
{Name: core.CodecH265},
},
},
{
Kind: core.KindAudio,
Direction: core.DirectionSendonly,
Codecs: []*core.Codec{
{Name: core.CodecAAC},
},
},
}
if v[0] == "" {
return medias // legacy
}
medias[1].Codecs = append(medias[1].Codecs,
&core.Codec{Name: core.CodecPCMA},
&core.Codec{Name: core.CodecPCMU},
&core.Codec{Name: core.CodecPCM},
&core.Codec{Name: core.CodecPCML},
)
if v[0] == "flac" {
return medias // modern browsers
}
medias[1].Codecs = append(medias[1].Codecs,
&core.Codec{Name: core.CodecOpus},
&core.Codec{Name: core.CodecMP3},
)
return medias // Chrome, FFmpeg, VLC
}
return core.ParseQuery(query)
}
func ParseCodecs(codecs string, parseAudio bool) (medias []*core.Media) {
var videos []*core.Codec
var audios []*core.Codec
for _, name := range strings.Split(codecs, ",") {
switch name {
case MimeH264:
codec := &core.Codec{Name: core.CodecH264}
videos = append(videos, codec)
case MimeH265:
codec := &core.Codec{Name: core.CodecH265}
videos = append(videos, codec)
case MimeAAC:
codec := &core.Codec{Name: core.CodecAAC}
audios = append(audios, codec)
case MimeFlac:
audios = append(audios,
&core.Codec{Name: core.CodecPCMA},
&core.Codec{Name: core.CodecPCMU},
&core.Codec{Name: core.CodecPCM},
&core.Codec{Name: core.CodecPCML},
)
case MimeOpus:
codec := &core.Codec{Name: core.CodecOpus}
audios = append(audios, codec)
}
}
if videos != nil {
media := &core.Media{
Kind: core.KindVideo,
Direction: core.DirectionSendonly,
Codecs: videos,
}
medias = append(medias, media)
}
if audios != nil && parseAudio {
media := &core.Media{
Kind: core.KindAudio,
Direction: core.DirectionSendonly,
Codecs: audios,
}
medias = append(medias, media)
}
return
}
// PatchVideoRotate - update video track transformation matrix.
// Rotation supported by many players and browsers (except Safari).
// Scale has low support and better not to use it.
// Supported only 0, 90, 180, 270 degrees.
func PatchVideoRotate(init []byte, degrees int) bool {
// search video atom
i := bytes.Index(init, []byte("vide"))
if i < 0 {
return false
}
// seek to video matrix position
i -= 4 + 3 + 1 + 8 + 32 + 8 + 4 + 4 + 4*9
// Rotation matrix:
// [ cos sin 0]
// [ -sin cos 0]
// [ 0 0 16384]
var cos, sin uint16
switch degrees {
case 0:
cos = 1
sin = 0
case 90:
cos = 0
sin = 1
case 180:
cos = 0xFFFF // -1
sin = 0
case 270:
cos = 0
sin = 0xFFFF // -1
default:
return false
}
binary.BigEndian.PutUint16(init[i:], cos)
binary.BigEndian.PutUint16(init[i+4:], sin)
binary.BigEndian.PutUint16(init[i+12:], -sin)
binary.BigEndian.PutUint16(init[i+16:], cos)
return true
}
// PatchVideoScale - update "Pixel Aspect Ratio" atom.
// Supported by many players and browsers (except Firefox).
// Supported only positive integers.
func PatchVideoScale(init []byte, scaleX, scaleY int) bool {
// search video atom
i := bytes.Index(init, []byte("pasp"))
if i < 0 {
return false
}
binary.BigEndian.PutUint32(init[i+4:], uint32(scaleX))
binary.BigEndian.PutUint32(init[i+8:], uint32(scaleY))
return true
}
+104
View File
@@ -0,0 +1,104 @@
package mp4
import (
"io"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264"
"github.com/AlexxIT/go2rtc/pkg/h265"
"github.com/pion/rtp"
)
type Keyframe struct {
core.Connection
wr *core.WriteBuffer
muxer *Muxer
}
// Deprecated: should be rewritten
func NewKeyframe(medias []*core.Media) *Keyframe {
if medias == nil {
medias = []*core.Media{
{
Kind: core.KindVideo,
Direction: core.DirectionSendonly,
Codecs: []*core.Codec{
{Name: core.CodecH264},
{Name: core.CodecH265},
},
},
}
}
wr := core.NewWriteBuffer(nil)
cons := &Keyframe{
Connection: core.Connection{
ID: core.NewID(),
FormatName: "mp4",
Transport: wr,
},
muxer: &Muxer{},
wr: wr,
}
cons.Medias = medias
return cons
}
func (c *Keyframe) AddTrack(media *core.Media, _ *core.Codec, track *core.Receiver) error {
c.muxer.AddTrack(track.Codec)
init, err := c.muxer.GetInit()
if err != nil {
return err
}
handler := core.NewSender(media, track.Codec)
switch track.Codec.Name {
case core.CodecH264:
handler.Handler = func(packet *rtp.Packet) {
if !h264.IsKeyframe(packet.Payload) {
return
}
// important to use Mutex because right fragment order
b := c.muxer.GetPayload(0, packet)
b = append(init, b...)
if n, err := c.wr.Write(b); err == nil {
c.Send += n
}
}
if track.Codec.IsRTP() {
handler.Handler = h264.RTPDepay(track.Codec, handler.Handler)
} else {
handler.Handler = h264.RepairAVCC(track.Codec, handler.Handler)
}
case core.CodecH265:
handler.Handler = func(packet *rtp.Packet) {
if !h265.IsKeyframe(packet.Payload) {
return
}
// important to use Mutex because right fragment order
b := c.muxer.GetPayload(0, packet)
b = append(init, b...)
if n, err := c.wr.Write(b); err == nil {
c.Send += n
}
}
if track.Codec.IsRTP() {
handler.Handler = h265.RTPDepay(track.Codec, handler.Handler)
}
}
handler.HandleRTP(track)
c.Senders = append(c.Senders, handler)
return nil
}
func (c *Keyframe) WriteTo(wr io.Writer) (int64, error) {
return c.wr.WriteTo(wr)
}
+45
View File
@@ -0,0 +1,45 @@
package mp4
import (
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264"
)
const (
MimeH264 = "avc1.640029"
MimeH265 = "hvc1.1.6.L153.B0"
MimeAAC = "mp4a.40.2"
MimeFlac = "flac"
MimeOpus = "opus"
)
func MimeCodecs(codecs []*core.Codec) string {
var s string
for i, codec := range codecs {
if i > 0 {
s += ","
}
switch codec.Name {
case core.CodecH264:
s += "avc1." + h264.GetProfileLevelID(codec.FmtpLine)
case core.CodecH265:
// H.265 profile=main level=5.1
// hvc1 - supported in Safari, hev1 - doesn't, both supported in Chrome
s += MimeH265
case core.CodecAAC:
s += MimeAAC
case core.CodecOpus:
s += MimeOpus
case core.CodecFLAC:
s += MimeFlac
}
}
return s
}
func ContentType(codecs []*core.Codec) string {
return `video/mp4; codecs="` + MimeCodecs(codecs) + `"`
}
+172
View File
@@ -0,0 +1,172 @@
package mp4
import (
"encoding/hex"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264"
"github.com/AlexxIT/go2rtc/pkg/h265"
"github.com/AlexxIT/go2rtc/pkg/iso"
"github.com/pion/rtp"
)
type Muxer struct {
index uint32
dts []uint64
pts []uint32
codecs []*core.Codec
}
func (m *Muxer) AddTrack(codec *core.Codec) {
m.dts = append(m.dts, 0)
m.pts = append(m.pts, 0)
m.codecs = append(m.codecs, codec)
}
func (m *Muxer) GetInit() ([]byte, error) {
mv := iso.NewMovie(1024)
mv.WriteFileType()
mv.StartAtom(iso.Moov)
mv.WriteMovieHeader()
for i, codec := range m.codecs {
switch codec.Name {
case core.CodecH264:
sps, pps := h264.GetParameterSet(codec.FmtpLine)
// some dummy SPS and PPS not a problem for MP4, but problem for HLS :(
if len(sps) == 0 {
sps = []byte{0x67, 0x42, 0x00, 0x0a, 0xf8, 0x41, 0xa2}
}
if len(pps) == 0 {
pps = []byte{0x68, 0xce, 0x38, 0x80}
}
var width, height uint16
if s := h264.DecodeSPS(sps); s != nil {
width = s.Width()
height = s.Height()
} else {
width = 1920
height = 1080
}
mv.WriteVideoTrack(
uint32(i+1), codec.Name, codec.ClockRate, width, height, h264.EncodeConfig(sps, pps),
)
case core.CodecH265:
vps, sps, pps := h265.GetParameterSet(codec.FmtpLine)
// some dummy SPS and PPS not a problem
if len(vps) == 0 {
vps = []byte{0x40, 0x01, 0x0c, 0x01, 0xff, 0xff, 0x01, 0x40, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x99, 0xac, 0x09}
}
if len(sps) == 0 {
sps = []byte{0x42, 0x01, 0x01, 0x01, 0x40, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x00, 0x03, 0x00, 0x99, 0xa0, 0x01, 0x40, 0x20, 0x05, 0xa1, 0xfe, 0x5a, 0xee, 0x46, 0xc1, 0xae, 0x55, 0x04}
}
if len(pps) == 0 {
pps = []byte{0x44, 0x01, 0xc0, 0x73, 0xc0, 0x4c, 0x90}
}
var width, height uint16
if s := h265.DecodeSPS(sps); s != nil {
width = s.Width()
height = s.Height()
} else {
width = 1920
height = 1080
}
mv.WriteVideoTrack(
uint32(i+1), codec.Name, codec.ClockRate, width, height, h265.EncodeConfig(vps, sps, pps),
)
case core.CodecAAC:
s := core.Between(codec.FmtpLine, "config=", ";")
b, err := hex.DecodeString(s)
if err != nil {
return nil, err
}
mv.WriteAudioTrack(
uint32(i+1), codec.Name, codec.ClockRate, uint16(codec.Channels), b,
)
case core.CodecOpus, core.CodecMP3, core.CodecPCMA, core.CodecPCMU, core.CodecFLAC:
mv.WriteAudioTrack(
uint32(i+1), codec.Name, codec.ClockRate, uint16(codec.Channels), nil,
)
}
}
mv.StartAtom(iso.MoovMvex)
for i := range m.codecs {
mv.WriteTrackExtend(uint32(i + 1))
}
mv.EndAtom() // MVEX
mv.EndAtom() // MOOV
return mv.Bytes(), nil
}
func (m *Muxer) Reset() {
m.index = 0
for i := range m.dts {
m.dts[i] = 0
m.pts[i] = 0
}
}
func (m *Muxer) GetPayload(trackID byte, packet *rtp.Packet) []byte {
codec := m.codecs[trackID]
m.index++
duration := packet.Timestamp - m.pts[trackID]
m.pts[trackID] = packet.Timestamp
// flags important for Apple Finder video preview
var flags uint32
switch codec.Name {
case core.CodecH264:
if h264.IsKeyframe(packet.Payload) {
flags = iso.SampleVideoIFrame
} else {
flags = iso.SampleVideoNonIFrame
}
case core.CodecH265:
if h265.IsKeyframe(packet.Payload) {
flags = iso.SampleVideoIFrame
} else {
flags = iso.SampleVideoNonIFrame
}
case core.CodecAAC:
duration = 1024 // important for Apple Finder and QuickTime
flags = iso.SampleAudio // not important?
default:
flags = iso.SampleAudio // important for FLAC on Android Telegram
}
// minumum duration important for MSE in Apple Safari
if duration == 0 || duration > codec.ClockRate {
duration = codec.ClockRate/1000 + 1
m.pts[trackID] += duration
}
size := len(packet.Payload)
mv := iso.NewMovie(1024 + size)
mv.WriteMovieFragment(
// ExtensionProfile - wrong place for CTS (supported by mpegts.Demuxer)
m.index, uint32(trackID+1), duration, uint32(size), flags, m.dts[trackID], uint32(packet.ExtensionProfile),
)
mv.WriteData(packet.Payload)
//log.Printf("[MP4] idx:%3d trk:%d dts:%6d cts:%4d dur:%5d time:%10d len:%5d", m.index, trackID+1, m.dts[trackID], packet.SSRC, duration, packet.Timestamp, len(packet.Payload))
m.dts[trackID] += uint64(duration)
return mv.Bytes()
}