From d203d75bdd8165112d4f9793cd556256f14983d6 Mon Sep 17 00:00:00 2001
From: Vanilagy <1696106+Vanilagy@users.noreply.github.com>
Date: Mon, 30 Dec 2024 23:47:42 +0100
Subject: [PATCH] Add MP3 support to ISOBMFF demuxer

---
 src/codec.ts                   |  6 +++-
 src/isobmff/isobmff-demuxer.ts | 51 +++++++++++++++++++++-------------
 2 files changed, 37 insertions(+), 20 deletions(-)

diff --git a/src/codec.ts b/src/codec.ts
index 5117df7..a2e4e85 100644
--- a/src/codec.ts
+++ b/src/codec.ts
@@ -13,7 +13,7 @@ import { SubtitleMetadata } from './subtitles';
 /** @public */
 export const VIDEO_CODECS = ['avc', 'hevc', 'vp8', 'vp9', 'av1'] as const;
 /** @public */
-export const AUDIO_CODECS = ['aac', 'opus'] as const; // TODO add the rest
+export const AUDIO_CODECS = ['aac', 'mp3', 'opus'] as const; // TODO add the rest
 /** @public */
 export const SUBTITLE_CODECS = ['webvtt'] as const; // TODO add the rest
 
@@ -353,6 +353,8 @@ export const buildAudioCodecString = (codec: AudioCodec, numberOfChannels: numbe
 
 		// Default to standard AAC-LC for higher sample rates
 		return 'mp4a.40.2'; // AAC-LC
+	} else if (codec === 'mp3') {
+		return 'mp3'; // Easy, this one
 	} else if (codec === 'opus') {
 		return 'opus'; // Easy, this one
 	} else if (codec === 'vorbis') {
@@ -367,6 +369,8 @@ export const extractAudioCodecString = (codec: AudioCodec, description: Uint8Arr
 	if (codec === 'aac') {
 		const audioSpecificConfig = parseAacAudioSpecificConfig(description);
 		return `mp4a.40.${audioSpecificConfig.objectType}`;
+	} else if (codec === 'mp3') {
+		return 'mp3';
 	} else if (codec === 'opus') {
 		return 'opus';
 	} else if (codec === 'vorbis') {
diff --git a/src/isobmff/isobmff-demuxer.ts b/src/isobmff/isobmff-demuxer.ts
index 4d775b1..0708ef9 100644
--- a/src/isobmff/isobmff-demuxer.ts
+++ b/src/isobmff/isobmff-demuxer.ts
@@ -613,7 +613,7 @@ export class IsobmffDemuxer extends Demuxer {
 						this.readContiguousBoxes(startPos + sampleBoxInfo.totalSize - this.isobmffReader.pos);
 					} else {
 						if (sampleBoxInfo.name === 'mp4a') {
-							track.info.codec = 'aac';
+							// We don't know the codec yet, need to read the esds box
 						} else if (sampleBoxInfo.name.toLowerCase() === 'opus') {
 							track.info.codec = 'opus';
 						} else {
@@ -811,12 +811,12 @@ export class IsobmffDemuxer extends Demuxer {
 
 			case 'esds': {
 				const track = this.currentTrack;
-				assert(track && track.info);
+				assert(track && track.info?.type === 'audio');
 
 				this.isobmffReader.pos += 4; // Version + flags
 
 				const tag = this.isobmffReader.readU8();
-				assert(tag === 0x03);
+				assert(tag === 0x03); // ES Descriptor
 
 				this.isobmffReader.readIsomVariableInteger(); // Length
 
@@ -839,24 +839,38 @@ export class IsobmffDemuxer extends Demuxer {
 				}
 
 				const decoderConfigTag = this.isobmffReader.readU8();
-				assert(decoderConfigTag === 0x04);
+				assert(decoderConfigTag === 0x04); // DecoderConfigDescriptor
 
-				this.isobmffReader.readIsomVariableInteger(); // Length
+				const decoderConfigDescriptorLength = this.isobmffReader.readIsomVariableInteger(); // Length
+
+				const payloadStart = this.isobmffReader.pos;
 
 				const objectTypeIndication = this.isobmffReader.readU8();
-				assert(objectTypeIndication === 0x40); // Assert it's MPEG-4 audio
+				if (objectTypeIndication === 0x40) {
+					track.info.codec = 'aac';
+				} else if (objectTypeIndication === 0x69 || objectTypeIndication === 0x6b) {
+					track.info.codec = 'mp3';
+				} else {
+					console.warn(
+						`Unsupported audio codec (objectTypeIndication ${objectTypeIndication}) - discarding track.`,
+					);
+				}
 
 				this.isobmffReader.pos += 1 + 3 + 4 + 4;
 
-				const decoderSpecificInfoTag = this.isobmffReader.readU8();
-				assert(decoderSpecificInfoTag === 0x05);
+				if (decoderConfigDescriptorLength > this.isobmffReader.pos - payloadStart) {
+					// There's a DecoderSpecificInfo at the end, let's read it
 
-				const decoderSpecificInfoLength = this.isobmffReader.readIsomVariableInteger();
+					const decoderSpecificInfoTag = this.isobmffReader.readU8();
+					assert(decoderSpecificInfoTag === 0x05); // DecoderSpecificInfo
 
-				track.info.codecDescription = this.isobmffReader.readRange(
-					this.isobmffReader.pos,
-					this.isobmffReader.pos + decoderSpecificInfoLength,
-				);
+					const decoderSpecificInfoLength = this.isobmffReader.readIsomVariableInteger();
+
+					track.info.codecDescription = this.isobmffReader.readRange(
+						this.isobmffReader.pos,
+						this.isobmffReader.pos + decoderSpecificInfoLength,
+					);
+				}
 			}; break;
 
 			case 'stts': {
@@ -2003,17 +2017,16 @@ const getSampleInfo = (sampleTable: SampleTable, sampleIndex: number): SampleInf
 		+ Math.floor((sampleIndex - chunkEntry.startSampleIndex) / chunkEntry.samplesPerChunk);
 	const chunkOffset = sampleTable.chunkOffsets[chunkIndex]!;
 
+	const startSampleIndexOfChunk = chunkEntry.startSampleIndex
+		+ (chunkIndex - chunkEntry.startChunkIndex) * chunkEntry.samplesPerChunk;
 	let chunkSize = 0;
-
 	let sampleOffset = chunkOffset;
+
 	if (sampleTable.sampleSizes.length === 1) {
-		sampleOffset += sampleSize * (sampleIndex - chunkEntry.startSampleIndex);
+		sampleOffset += sampleSize * (sampleIndex - startSampleIndexOfChunk);
 		chunkSize += sampleSize * chunkEntry.samplesPerChunk;
 	} else {
-		const startSampleIndex = chunkEntry.startSampleIndex
-			+ (chunkIndex - chunkEntry.startChunkIndex) * chunkEntry.samplesPerChunk;
-
-		for (let i = startSampleIndex; i < startSampleIndex + chunkEntry.samplesPerChunk; i++) {
+		for (let i = startSampleIndexOfChunk; i < startSampleIndexOfChunk + chunkEntry.samplesPerChunk; i++) {
 			const sampleSize = sampleTable.sampleSizes[i]!;
 
 			if (i < sampleIndex) {