MP4Generator: Generate .mp3 box for mp3 audio track in InitSegment

This commit makes mp3 audio playback works well in FireFox.
leicc520 · Mar 7, 2017 · 8dbea65 · 8dbea65
1 parent e1db184
commit 8dbea65
Show file tree

Hide file tree

Showing 2 changed files with 43 additions and 17 deletions.
diff --git a/src/demux/flv-demuxer.js b/src/demux/flv-demuxer.js
@@ -475,7 +475,6 @@ class FLVDemuxer {
                 meta.channelCount = misc.channelCount;
                 meta.codec = misc.codec;
                 meta.config = misc.config;
-                meta.esdsSubType = 0x40;  // Audio 14496-3 AAC Main
                 // The decode result of an aac sample is 1024 PCM samples
                 meta.refSampleDuration = Math.floor(1024 / meta.audioSampleRate * meta.timescale);
                 Log.v(this.TAG, 'Parsed AudioSpecificConfig');
@@ -516,16 +515,17 @@ class FLVDemuxer {
             }
         } else if (soundFormat === 2) {  // MP3
             if (!meta.codec) {
+                // We need metadata for mp3 audio track, extract info from frame header
                 let misc = this._parseMP3AudioData(arrayBuffer, dataOffset + 1, dataSize - 1, true);
                 if (misc == undefined) {
                     return;
                 }
                 meta.audioSampleRate = misc.samplingRate;
                 meta.channelConfig = misc.channelCount;
                 meta.codec = misc.codec;
-                meta.esdsSubType = 0x6B;
+                // The decode result of an mp3 sample is 1152 PCM samples
                 meta.refSampleDuration = Math.floor(1152 / meta.audioSampleRate * meta.timescale);
-                Log.v(this.TAG, 'Parsed MP3 frame header');
+                Log.v(this.TAG, 'Parsed MPEG Audio Frame Header');
 
                 this._audioInitialMetadataDispatched = true;
                 this._onTrackMetadata('audio', meta);
@@ -534,7 +534,7 @@ class FLVDemuxer {
                 mi.audioCodec = meta.codec;
                 mi.audioSampleRate = meta.audioSampleRate;
                 mi.audioChannelCount = meta.channelCount;
-                mi.audioBitrate = misc.bitRate;
+                mi.audioDataRate = misc.bitRate;
                 if (mi.hasVideo) {
                     if (mi.videoCodec != null) {
                         mi.mimeType = 'video/x-flv; codecs="' + mi.videoCodec + ',' + mi.audioCodec + '"';
@@ -545,16 +545,17 @@ class FLVDemuxer {
                 if (mi.isComplete()) {
                     this._onMediaInfo(mi);
                 }
-            } else {
-                let data = this._parseMP3AudioData(arrayBuffer, dataOffset + 1, dataSize - 1, false);
-                if (data == undefined) {
-                    return;
-                }
-                let dts = this._timestampBase + tagTimestamp;
-                let mp3Sample = {unit: data, dts: dts, pts: dts};
-                track.samples.push(mp3Sample);
-                track.length += data.length;
             }
+
+            // This packet is always a valid audio packet, extract it
+            let data = this._parseMP3AudioData(arrayBuffer, dataOffset + 1, dataSize - 1, false);
+            if (data == undefined) {
+                return;
+            }
+            let dts = this._timestampBase + tagTimestamp;
+            let mp3Sample = {unit: data, dts: dts, pts: dts};
+            track.samples.push(mp3Sample);
+            track.length += data.length;
         }
     }
 
@@ -706,6 +707,8 @@ class FLVDemuxer {
             let bit_rate = 0;
             let object_type = 34;  // Layer-3, listed in MPEG-4 Audio Object Types
 
+            let codec = 'mp3';
+
             switch (ver) {
                 case 0:  // MPEG 2.5
                     sample_rate = this._mpegAudioV25SampleRateTable[sampling_freq_index];
@@ -743,7 +746,7 @@ class FLVDemuxer {
                 bitRate: bit_rate,
                 samplingRate: sample_rate,
                 channelCount: channel_count,
-                codec: 'mp4a.40.' + object_type
+                codec: codec
             };
         } else {
             result = array;

diff --git a/src/remux/mp4-generator.js b/src/remux/mp4-generator.js
@@ -30,7 +30,7 @@ class MP4 {
             stco: [], stsc: [], stsd: [], stsz: [],
             stts: [], tfdt: [], tfhd: [], traf: [],
             trak: [], trun: [], trex: [], tkhd: [],
-            vmhd: [], smhd: []
+            vmhd: [], smhd: [], '.mp3': []
         };
 
         for (let name in MP4.types) {
@@ -318,12 +318,36 @@ class MP4 {
     // Sample description box
     static stsd(meta) {
         if (meta.type === 'audio') {
+            if (meta.codec === 'mp3') {
+                return MP4.box(MP4.types.stsd, MP4.constants.STSD_PREFIX, MP4.mp3(meta));
+            }
+            // else: aac -> mp4a
             return MP4.box(MP4.types.stsd, MP4.constants.STSD_PREFIX, MP4.mp4a(meta));
         } else {
             return MP4.box(MP4.types.stsd, MP4.constants.STSD_PREFIX, MP4.avc1(meta));
         }
     }
 
+    static mp3(meta) {
+        let channelCount = meta.channelCount;
+        let sampleRate = meta.audioSampleRate;
+
+        let data = new Uint8Array([
+            0x00, 0x00, 0x00, 0x00,  // reserved(4)
+            0x00, 0x00, 0x00, 0x01,  // reserved(2) + data_reference_index(2)
+            0x00, 0x00, 0x00, 0x00,  // reserved: 2 * 4 bytes
+            0x00, 0x00, 0x00, 0x00,
+            0x00, channelCount,      // channelCount(2)
+            0x00, 0x10,              // sampleSize(2)
+            0x00, 0x00, 0x00, 0x00,  // reserved(4)
+            (sampleRate >>> 8) & 0xFF,  // Audio sample rate
+            (sampleRate) & 0xFF,
+            0x00, 0x00
+        ]);
+
+        return MP4.box(MP4.types['.mp3'], data);
+    }
+
     static mp4a(meta) {
         let channelCount = meta.channelCount;
         let sampleRate = meta.audioSampleRate;
@@ -345,7 +369,6 @@ class MP4 {
     }
 
     static esds(meta) {
-        let esdsSubType = meta.esdsSubType;
         let config = meta.config || [];
         let configSize = config.length;
         let data = new Uint8Array([
@@ -358,7 +381,7 @@ class MP4 {
 
             0x04,                    // descriptor_type
             0x0F + configSize,       // length
-            esdsSubType,             // codec
+            0x40,                    // codec: mpeg4_audio
             0x15,                    // stream_type: Audio
             0x00, 0x00, 0x00,        // buffer_size
             0x00, 0x00, 0x00, 0x00,  // maxBitrate