From 1612f9c81e3714d6c431f4e078b628dfed327169 Mon Sep 17 00:00:00 2001
From: Alexey Khit <alexey.khit@gmail.com>
Date: Thu, 25 May 2023 06:49:04 +0300
Subject: [PATCH] Fix AAC inside MP4

---
 pkg/aac/rtp.go   | 32 ++++++++++++++++++++++++--------
 pkg/iso/atoms.go |  1 +
 pkg/mp4/muxer.go | 22 +++++++++++++---------
 3 files changed, 38 insertions(+), 17 deletions(-)

diff --git a/pkg/aac/rtp.go b/pkg/aac/rtp.go
index ba3dfce9..ec92ebad 100644
--- a/pkg/aac/rtp.go
+++ b/pkg/aac/rtp.go
@@ -9,6 +9,8 @@ import (
 const RTPPacketVersionAAC = 0
 
 func RTPDepay(handler core.HandlerFunc) core.HandlerFunc {
+	var timestamp uint32
+
 	return func(packet *rtp.Packet) {
 		// support ONLY 2 bytes header size!
 		// streamtype=5;profile-level-id=1;mode=AAC-hbr;sizelength=13;indexlength=3;indexdeltalength=3;config=1408
@@ -16,15 +18,29 @@ func RTPDepay(handler core.HandlerFunc) core.HandlerFunc {
 
 		//log.Printf("[RTP/AAC] units: %d, size: %4d, ts: %10d, %t", headersSize/2, len(packet.Payload), packet.Timestamp, packet.Marker)
 
-		data := packet.Payload[2+headersSize:]
-		if IsADTS(data) {
-			data = data[7:]
-		}
+		headers := packet.Payload[2 : 2+headersSize]
+		units := packet.Payload[2+headersSize:]
 
-		clone := *packet
-		clone.Version = RTPPacketVersionAAC
-		clone.Payload = data
-		handler(&clone)
+		for len(headers) > 0 {
+			unitSize := binary.BigEndian.Uint16(headers) >> 3
+
+			unit := units[:unitSize]
+
+			headers = headers[2:]
+			units = units[unitSize:]
+
+			timestamp += 1024
+
+			clone := *packet
+			clone.Version = RTPPacketVersionAAC
+			clone.Timestamp = timestamp
+			if IsADTS(unit) {
+				clone.Payload = unit[7:]
+			} else {
+				clone.Payload = unit
+			}
+			handler(&clone)
+		}
 	}
 }
 
diff --git a/pkg/iso/atoms.go b/pkg/iso/atoms.go
index 6a4c9fe7..174f16ea 100644
--- a/pkg/iso/atoms.go
+++ b/pkg/iso/atoms.go
@@ -40,6 +40,7 @@ const (
 	SampleVideoIFrame    = sampleDependsOn2
 	SampleVideoNonIFrame = sampleDependsOn1 | sampleIsNonSync
 	SampleAudio          = sampleIsNonSync
+	SampleAudioAAC       = sampleDependsOn2
 )
 
 func (m *Movie) WriteFileType() {
diff --git a/pkg/mp4/muxer.go b/pkg/mp4/muxer.go
index b1ec37f4..8f9afc4c 100644
--- a/pkg/mp4/muxer.go
+++ b/pkg/mp4/muxer.go
@@ -151,19 +151,14 @@ func (m *Muxer) Reset() {
 func (m *Muxer) Marshal(trackID byte, packet *rtp.Packet) []byte {
 	codec := m.codecs[trackID]
 
+	m.fragIndex++
+
 	duration := packet.Timestamp - m.pts[trackID]
 	m.pts[trackID] = packet.Timestamp
 
-	// 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)
-
 	// flags important for Apple Finder video preview
 	var flags uint32
+
 	switch codec.Name {
 	case core.CodecH264:
 		if h264.IsKeyframe(packet.Payload) {
@@ -177,11 +172,20 @@ func (m *Muxer) Marshal(trackID byte, packet *rtp.Packet) []byte {
 		} else {
 			flags = iso.SampleVideoNonIFrame
 		}
+	case core.CodecAAC:
+		duration = 1024            // important for Apple Finder and QuickTime
+		flags = iso.SampleAudioAAC // not important
 	default:
 		flags = iso.SampleAudio // not important
 	}
 
-	m.fragIndex++
+	// 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(