infer object, sample rate, clock of MPEG4AudioLATM from config (#255)

go.mod

@@ -4,7 +4,7 @@ go 1.18
 
 require (
 	github.com/asticode/go-astits v1.11.0
-	github.com/bluenviron/mediacommon v0.2.1
+	github.com/bluenviron/mediacommon v0.3.1
 	github.com/google/uuid v1.3.0
 	github.com/pion/rtcp v1.2.10
 	github.com/pion/rtp v0.0.0-20230107162714-c3ea6851e25b

go.sum

@@ -2,8 +2,8 @@ github.com/asticode/go-astikit v0.30.0 h1:DkBkRQRIxYcknlaU7W7ksNfn4gMFsB0tqMJflx
 github.com/asticode/go-astikit v0.30.0/go.mod h1:h4ly7idim1tNhaVkdVBeXQZEE3L0xblP7fCWbgwipF0=
 github.com/asticode/go-astits v1.11.0 h1:GTHUXht0ZXAJXsVbsLIcyfHr1Bchi4QQwMARw2ZWAng=
 github.com/asticode/go-astits v1.11.0/go.mod h1:QSHmknZ51pf6KJdHKZHJTLlMegIrhega3LPWz3ND/iI=
-github.com/bluenviron/mediacommon v0.2.1 h1:NkbbXwjMqspatQwSyGDYqxg14ftxJSjNsnPNocUxScw=
+github.com/bluenviron/mediacommon v0.3.1 h1:C4okNqyN1Mg5CVGcGKk2tEk9Uj2hHZusHV7nqdjn1Lk=
-github.com/bluenviron/mediacommon v0.2.1/go.mod h1:t0dqPsWUTchyvib0MhixIwXEgvDX4V9G+I0GzWLQRb8=
+github.com/bluenviron/mediacommon v0.3.1/go.mod h1:t0dqPsWUTchyvib0MhixIwXEgvDX4V9G+I0GzWLQRb8=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=

pkg/formats/format_test.go

@@ -233,7 +233,40 @@ var casesFormat = []struct {
 		},
 	},
 	{
-		"audio aac lc latm",
+		"audio aac he-aac v2 ps",
+		"audio",
+		96,
+		"mpeg4-generic/48000/2",
+		map[string]string{
+			"streamtype":       "5",
+			"profile-level-id": "48",
+			"mode":             "AAC-hbr",
+			"config":           "eb098800",
+			"sizelength":       "13",
+		},
+		&MPEG4Audio{
+			PayloadTyp:     96,
+			ProfileLevelID: 48,
+			Config: &mpeg4audio.Config{
+				Type:                2,
+				ExtensionType:       29,
+				ExtensionSampleRate: 48000,
+				SampleRate:          24000,
+				ChannelCount:        1,
+			},
+			SizeLength: 13,
+		},
+		"mpeg4-generic/48000/2",
+		map[string]string{
+			"streamtype":       "5",
+			"profile-level-id": "48",
+			"mode":             "AAC-hbr",
+			"config":           "eb098800",
+			"sizelength":       "13",
+		},
+	},
+	{
+		"audio aac latm lc",
 		"audio",
 		96,
 		"MP4A-LATM/24000/2",
@@ -246,13 +279,21 @@ var casesFormat = []struct {
 		},
 		&MPEG4AudioLATM{
 			PayloadTyp:     96,
-			SampleRate:     24000,
-			Channels:       2,
 			ProfileLevelID: 1,
 			Bitrate:        intPtr(64000),
 			CPresent:       boolPtr(false),
-			Object:         2,
+			Config: &mpeg4audio.StreamMuxConfig{
-			Config:         []byte{0x40, 0x00, 0x26, 0x20, 0x3f, 0xc0},
+				Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+					Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+						AudioSpecificConfig: &mpeg4audio.Config{
+							Type:         2,
+							SampleRate:   24000,
+							ChannelCount: 2,
+						},
+						LatmBufferFullness: 255,
+					}},
+				}},
+			},
 		},
 		"MP4A-LATM/24000/2",
 		map[string]string{
@@ -264,7 +305,7 @@ var casesFormat = []struct {
 		},
 	},
 	{
-		"audio aac v2 latm",
+		"audio aac latm he-aac v2",
 		"audio",
 		110,
 		"MP4A-LATM/24000/1",
@@ -277,13 +318,21 @@ var casesFormat = []struct {
 		},
 		&MPEG4AudioLATM{
 			PayloadTyp:     110,
-			SampleRate:     24000,
-			Channels:       1,
 			ProfileLevelID: 15,
 			CPresent:       boolPtr(false),
-			Object:         2,
 			SBREnabled:     boolPtr(true),
-			Config:         []byte{0x40, 0x00, 0x26, 0x10, 0x3f, 0xc0},
+			Config: &mpeg4audio.StreamMuxConfig{
+				Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+					Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+						AudioSpecificConfig: &mpeg4audio.Config{
+							Type:         2,
+							SampleRate:   24000,
+							ChannelCount: 1,
+						},
+						LatmBufferFullness: 255,
+					}},
+				}},
+			},
 		},
 		"MP4A-LATM/24000/1",
 		map[string]string{
@@ -294,6 +343,89 @@ var casesFormat = []struct {
 			"SBR-enabled":      "1",
 		},
 	},
+	{
+		"audio aac latm hierarchical sbr",
+		"audio",
+		110,
+		"MP4A-LATM/48000/2",
+		map[string]string{
+			"profile-level-id": "44",
+			"bitrate":          "64000",
+			"cpresent":         "0",
+			"config":           "40005623101fe0",
+			"sbr-enabled":      "1",
+		},
+		&MPEG4AudioLATM{
+			PayloadTyp:     110,
+			ProfileLevelID: 44,
+			CPresent:       boolPtr(false),
+			SBREnabled:     boolPtr(true),
+			Bitrate:        intPtr(64000),
+			Config: &mpeg4audio.StreamMuxConfig{
+				Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+					Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+						AudioSpecificConfig: &mpeg4audio.Config{
+							Type:                2,
+							ExtensionType:       5,
+							ExtensionSampleRate: 48000,
+							SampleRate:          24000,
+							ChannelCount:        2,
+						},
+						LatmBufferFullness: 255,
+					}},
+				}},
+			},
+		},
+		"MP4A-LATM/48000/2",
+		map[string]string{
+			"profile-level-id": "44",
+			"object":           "2",
+			"cpresent":         "0",
+			"config":           "40005623101fe0",
+			"SBR-enabled":      "1",
+			"bitrate":          "64000",
+		},
+	},
+	{
+		"audio aac latm hierarchical ps",
+		"audio",
+		110,
+		"MP4A-LATM/48000/2",
+		map[string]string{
+			"profile-level-id": "48",
+			"bitrate":          "64000",
+			"cpresent":         "0",
+			"config":           "4001d613101fe0",
+		},
+		&MPEG4AudioLATM{
+			PayloadTyp:     110,
+			ProfileLevelID: 48,
+			Bitrate:        intPtr(64000),
+			CPresent:       boolPtr(false),
+			Config: &mpeg4audio.StreamMuxConfig{
+				Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+					Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+						AudioSpecificConfig: &mpeg4audio.Config{
+							Type:                2,
+							ExtensionType:       29,
+							ExtensionSampleRate: 48000,
+							SampleRate:          24000,
+							ChannelCount:        1,
+						},
+						LatmBufferFullness: 255,
+					}},
+				}},
+			},
+		},
+		"MP4A-LATM/48000/2",
+		map[string]string{
+			"profile-level-id": "48",
+			"object":           "2",
+			"cpresent":         "0",
+			"config":           "4001d613101fe0",
+			"bitrate":          "64000",
+		},
+	},
 	{
 		"audio aac latm no channels",
 		"audio",
@@ -305,19 +437,27 @@ var casesFormat = []struct {
 		},
 		&MPEG4AudioLATM{
 			PayloadTyp:     110,
-			SampleRate:     48000,
-			Channels:       1,
 			ProfileLevelID: 30,
 			CPresent:       boolPtr(false),
-			Object:         2,
+			Config: &mpeg4audio.StreamMuxConfig{
-			Config:         []byte{0x40, 0x00, 0x23, 0x10},
+				Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+					Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+						AudioSpecificConfig: &mpeg4audio.Config{
+							Type:         2,
+							SampleRate:   48000,
+							ChannelCount: 1,
+						},
+						LatmBufferFullness: 255,
+					}},
+				}},
+			},
 		},
 		"MP4A-LATM/48000/1",
 		map[string]string{
 			"profile-level-id": "30",
 			"object":           "2",
 			"cpresent":         "0",
-			"config":           "40002310",
+			"config":           "400023103fc0",
 		},
 	},
 	{

pkg/formats/mpeg4_audio_generic.go

@@ -120,8 +120,13 @@ func (f *MPEG4AudioGeneric) RTPMap() string {
 		sampleRate = f.Config.ExtensionSampleRate
 	}
 
+	channelCount := f.Config.ChannelCount
+	if f.Config.ExtensionType == mpeg4audio.ObjectTypePS {
+		channelCount = 2
+	}
+
 	return "mpeg4-generic/" + strconv.FormatInt(int64(sampleRate), 10) +
-		"/" + strconv.FormatInt(int64(f.Config.ChannelCount), 10)
+		"/" + strconv.FormatInt(int64(channelCount), 10)
 }
 
 // FMTP implements Format.

pkg/formats/mpeg4_audio_latm.go

@@ -4,7 +4,6 @@ import (
 	"encoding/hex"
 	"fmt"
 	"strconv"
-	"strings"
 
 	"github.com/bluenviron/mediacommon/pkg/codecs/mpeg4audio"
 	"github.com/pion/rtp"
@@ -14,13 +13,10 @@ import (
 // Specification: https://datatracker.ietf.org/doc/html/rfc6416#section-7.3
 type MPEG4AudioLATM struct {
 	PayloadTyp     uint8
-	SampleRate     int
-	Channels       int
 	ProfileLevelID int
 	Bitrate        *int
-	Object         mpeg4audio.ObjectType
 	CPresent       *bool
-	Config         []byte
+	Config         *mpeg4audio.StreamMuxConfig
 	SBREnabled     *bool
 }
 
@@ -28,24 +24,6 @@ func (f *MPEG4AudioLATM) unmarshal(
 	payloadType uint8, clock string, codec string,
 	rtpmap string, fmtp map[string]string,
 ) error {
-	tmp := strings.SplitN(clock, "/", 2)
-
-	tmp1, err := strconv.ParseInt(tmp[0], 10, 64)
-	if err != nil {
-		return err
-	}
-	f.SampleRate = int(tmp1)
-
-	if len(tmp) >= 2 {
-		tmp2, err := strconv.ParseInt(tmp[1], 10, 64)
-		if err != nil {
-			return err
-		}
-		f.Channels = int(tmp2)
-	} else {
-		f.Channels = 1
-	}
-
 	f.PayloadTyp = payloadType
 	f.ProfileLevelID = 30 // default value defined by specification
 
@@ -68,22 +46,6 @@ func (f *MPEG4AudioLATM) unmarshal(
 			v := int(tmp)
 			f.Bitrate = &v
 
-		case "object":
-			tmp, err := strconv.ParseInt(val, 10, 64)
-			if err != nil {
-				return fmt.Errorf("invalid object type: %v", val)
-			}
-
-			f.Object = mpeg4audio.ObjectType(int(tmp))
-
-			switch f.Object {
-			case mpeg4audio.ObjectTypeAACLC,
-				mpeg4audio.ObjectTypeSBR,
-				mpeg4audio.ObjectTypePS:
-			default:
-				return fmt.Errorf("unsupported object type: %d", f.Object)
-			}
-
 		case "cpresent":
 			tmp, err := strconv.ParseInt(val, 10, 64)
 			if err != nil {
@@ -94,12 +56,17 @@ func (f *MPEG4AudioLATM) unmarshal(
 			f.CPresent = &v
 
 		case "config":
-			var err error
+			enc, err := hex.DecodeString(val)
-			f.Config, err = hex.DecodeString(val)
 			if err != nil {
 				return fmt.Errorf("invalid AAC config: %v", val)
 			}
 
+			f.Config = &mpeg4audio.StreamMuxConfig{}
+			err = f.Config.Unmarshal(enc)
+			if err != nil {
+				return fmt.Errorf("invalid AAC config: %v", err)
+			}
+
 		case "sbr-enabled":
 			tmp, err := strconv.ParseInt(val, 10, 64)
 			if err != nil {
@@ -111,10 +78,6 @@ func (f *MPEG4AudioLATM) unmarshal(
 		}
 	}
 
-	if f.Object == 0 {
-		f.Object = 2
-	}
-
 	if f.Config == nil {
 		return fmt.Errorf("config is missing")
 	}
@@ -129,7 +92,7 @@ func (f *MPEG4AudioLATM) String() string {
 
 // ClockRate implements Format.
 func (f *MPEG4AudioLATM) ClockRate() int {
-	return f.SampleRate
+	return f.Config.Programs[0].Layers[0].AudioSpecificConfig.SampleRate
 }
 
 // PayloadType implements Format.
@@ -139,16 +102,33 @@ func (f *MPEG4AudioLATM) PayloadType() uint8 {
 
 // RTPMap implements Format.
 func (f *MPEG4AudioLATM) RTPMap() string {
-	return "MP4A-LATM/" + strconv.FormatInt(int64(f.SampleRate), 10) +
+	aoc := f.Config.Programs[0].Layers[0].AudioSpecificConfig
-		"/" + strconv.FormatInt(int64(f.Channels), 10)
+
+	sampleRate := aoc.SampleRate
+	if aoc.ExtensionSampleRate != 0 {
+		sampleRate = aoc.ExtensionSampleRate
+	}
+
+	channelCount := aoc.ChannelCount
+	if aoc.ExtensionType == mpeg4audio.ObjectTypePS {
+		channelCount = 2
+	}
+
+	return "MP4A-LATM/" + strconv.FormatInt(int64(sampleRate), 10) +
+		"/" + strconv.FormatInt(int64(channelCount), 10)
 }
 
 // FMTP implements Format.
 func (f *MPEG4AudioLATM) FMTP() map[string]string {
+	enc, err := f.Config.Marshal()
+	if err != nil {
+		return nil
+	}
+
 	fmtp := map[string]string{
 		"profile-level-id": strconv.FormatInt(int64(f.ProfileLevelID), 10),
-		"config":           hex.EncodeToString(f.Config),
+		"config":           hex.EncodeToString(enc),
-		"object":           strconv.FormatInt(int64(f.Object), 10),
+		"object":           strconv.FormatInt(int64(f.Config.Programs[0].Layers[0].AudioSpecificConfig.Type), 10),
 	}
 
 	if f.Bitrate != nil {

pkg/formats/mpeg4_audio_latm_test.go

@@ -3,6 +3,7 @@ package formats
 import (
 	"testing"
 
+	"github.com/bluenviron/mediacommon/pkg/codecs/mpeg4audio"
 	"github.com/pion/rtp"
 	"github.com/stretchr/testify/require"
 )
@@ -10,13 +11,21 @@ import (
 func TestMPEG4AudioLATMAttributes(t *testing.T) {
 	format := &MPEG4AudioLATM{
 		PayloadTyp:     96,
-		SampleRate:     48000,
-		Channels:       2,
-		Object:         2,
 		ProfileLevelID: 1,
-		Config:         []byte{0x01, 0x02, 0x03},
+		Config: &mpeg4audio.StreamMuxConfig{
+			Programs: []*mpeg4audio.StreamMuxConfigProgram{{
+				Layers: []*mpeg4audio.StreamMuxConfigLayer{{
+					AudioSpecificConfig: &mpeg4audio.Config{
+						Type:         2,
+						SampleRate:   44100,
+						ChannelCount: 2,
+					},
+					LatmBufferFullness: 255,
+				}},
+			}},
+		},
 	}
 	require.Equal(t, "MPEG4-audio-latm", format.String())
-	require.Equal(t, 48000, format.ClockRate())
+	require.Equal(t, 44100, format.ClockRate())
 	require.Equal(t, true, format.PTSEqualsDTS(&rtp.Packet{}))
 }