Handle implicit HE-AAC for AAC Main with low sample rate

Fixes #7667

src/demux/audio/adts.ts

@@ -13,7 +13,7 @@ import type {
 } from '../../types/demuxer';
 
 type AudioConfig = {
-  config: [number, number];
+  config: [number, number] | [number, number, number, number];
   samplerate: number;
   channelCount: number;
   codec: string;
@@ -52,7 +52,6 @@ export function getAudioConfig(
   // MPEG-4 Audio Object Type (profile_ObjectType+1)
   const adtsObjectType = ((byte2 >> 6) & 0x3) + 1;
   const channelCount = ((data[offset + 3] >> 6) & 0x3) | ((byte2 & 1) << 2);
-  const codec = 'mp4a.40.' + adtsObjectType;
   /* refer to http://wiki.multimedia.cx/index.php?title=MPEG-4_Audio#Audio_Specific_Config
       ISO/IEC 14496-3 - Table 1.13 — Syntax of AudioSpecificConfig()
     Audio Profile / Audio Object Type
@@ -95,12 +94,31 @@ export function getAudioConfig(
     // multiply frequency by 2 (see table above, equivalent to substract 3)
     aacSampleIndex -= 3;
   }
-  const config: [number, number] = [
-    (adtsObjectType << 3) | ((aacSampleIndex & 0x0e) >> 1),
-    ((aacSampleIndex & 0x01) << 7) | (channelCount << 3),
-  ];
+  let config: [number, number] | [number, number, number, number];
+  const samplingIndexHigh = (aacSampleIndex & 0x0e) >> 1;
+  const samplingIndexLowAndChannels =
+    ((aacSampleIndex & 0x01) << 7) | (channelCount << 3);
+  let adtsOverride = 0;
+  if (adtsObjectType === 1 && adtsSamplingIndex >= 6) {
+    // Handle implicit HE-AAC by adding explicit SBR signaling
+    adtsOverride = 5;
+    const adtsExtensionSamplingIndex = adtsSamplingIndex - 3;
+    config = [
+      (adtsOverride << 3) | samplingIndexHigh,
+      samplingIndexLowAndChannels | ((adtsExtensionSamplingIndex & 0x0e) >> 1),
+      ((adtsExtensionSamplingIndex & 0x01) << 7) | (2 << 2),
+      0,
+    ];
+  } else {
+    config = [
+      (adtsObjectType << 3) | samplingIndexHigh,
+      samplingIndexLowAndChannels,
+    ];
+  }
+  const codec = 'mp4a.40.' + (adtsOverride || adtsObjectType);
+
   logger.log(
-    `manifest codec:${manifestCodec}, parsed codec:${codec}, channels:${channelCount}, rate:${samplerate} (ADTS object type:${adtsObjectType} sampling index:${adtsSamplingIndex})`,
+    `manifest codec:${manifestCodec}, parsed codec:${codec}, channels:${channelCount}, rate:${samplerate} (ADTS object type:${adtsObjectType} sampling index:${adtsSamplingIndex} esds config:[${config.join(',')}])`,
   );
   return {
     config,

src/remux/mp4-generator.ts

@@ -695,7 +695,8 @@ class MP4 {
   }
 
   static esds(track: DemuxedAudioTrack) {
-    const config = track.config as [number, number];
+    const config = track.config!;
+    const len = config.length;
     return new Uint8Array([
       0x00, // version 0
       0x00,
@@ -703,7 +704,7 @@ class MP4 {
       0x00, // flags
 
       0x03, // descriptor_type
-      0x19, // length
+      0x17 + len, // length
 
       0x00,
       0x01, // es_id
@@ -711,7 +712,7 @@ class MP4 {
       0x00, // stream_priority
 
       0x04, // descriptor_type
-      0x11, // length
+      0x0f + len, // length
       0x40, // codec : mpeg4_audio
       0x15, // stream_type
       0x00,
@@ -727,7 +728,7 @@ class MP4 {
       0x00, // avgBitrate
 
       0x05, // descriptor_type
-      0x02, // length
+      len, // length
       ...config,
       0x06,
       0x01,

tests/unit/demuxer/adts.js

@@ -50,24 +50,43 @@ describe('getAudioConfig', function () {
     });
   });
 
-  it('should return audio config for 11025Hz', function () {
+  it('should treat AAC Main (object type 1) with low sample rate as implicit HE-AAC (SBR)', function () {
     const observer = new EventEmitter();
     const data = new Uint8Array(new ArrayBuffer(4));
     data[0] = 0xff;
     data[1] = 0xf0; // ID = 0 (MPEG-4), layer = 00, protection_absent = 0
-    data[2] = 0x28; // sampling_frequency_index = 10
+    data[2] = 0x28; // profile = 0 (AAC Main), sampling_frequency_index = 10 (11025Hz)
 
     const result = getAudioConfig(observer, data, 0, 'mp4a.40.29');
     expect(result, JSON.stringify(result)).to.deep.equal({
-      config: [13, 0],
+      config: [45, 3, 136, 0],
       samplerate: 11025,
       channelCount: 0,
-      codec: 'mp4a.40.1',
-      parsedCodec: 'mp4a.40.1',
+      codec: 'mp4a.40.5',
+      parsedCodec: 'mp4a.40.5',
       manifestCodec: 'mp4a.40.29',
     });
   });
 
+  it('should treat AAC Main at sampling index 6 (24kHz) as implicit HE-AAC', function () {
+    const observer = new EventEmitter();
+    const data = new Uint8Array(new ArrayBuffer(4));
+    data[0] = 0xff;
+    data[1] = 0xf0; // ID = 0 (MPEG-4), layer = 00, protection_absent = 0
+    data[2] = 0x18; // profile = 0 (AAC Main), sampling_frequency_index = 6 (24000Hz)
+    data[3] = 0x80; // channel_configuration = 2 (stereo)
+
+    const result = getAudioConfig(observer, data, 0, undefined);
+    expect(result, JSON.stringify(result)).to.deep.equal({
+      config: [43, 17, 136, 0],
+      samplerate: 24000,
+      channelCount: 2,
+      codec: 'mp4a.40.5',
+      parsedCodec: 'mp4a.40.5',
+      manifestCodec: undefined,
+    });
+  });
+
   it('should return audio config if there is no audio codec', function () {
     const observer = new EventEmitter();
     const data = new Uint8Array(new ArrayBuffer(4));
@@ -106,7 +125,7 @@ describe('getAudioConfig', function () {
     });
   });
 
-  it('should return audio config if there is no audio codec and freq is high enough', function () {
+  it('should keep AAC Main (object type 1) with high sample rate as mp4a.40.1', function () {
     const observer = new EventEmitter();
     const data = new Uint8Array(new ArrayBuffer(4));
     data[0] = 0xff;
@@ -272,7 +291,7 @@ describe('initTrackConfig', function () {
     });
   });
 
-  it('should call `getAudioConfig` and change track if track.samplerate is undefined', function () {
+  it('should set track config with implicit HE-AAC for AAC Main with low sample rate', function () {
     const track = {};
     const observer = {
       trigger: sinon.spy(),
@@ -285,11 +304,11 @@ describe('initTrackConfig', function () {
     initTrackConfig(track, observer, data, 0, 'mp4a.40.29');
 
     expect(track).to.deep.equal({
-      config: [13, 0],
+      config: [45, 3, 136, 0],
       samplerate: 11025,
       channelCount: 0,
-      codec: 'mp4a.40.1',
-      parsedCodec: 'mp4a.40.1',
+      codec: 'mp4a.40.5',
+      parsedCodec: 'mp4a.40.5',
       manifestCodec: 'mp4a.40.29',
     });
   });