Add support for ID3 metadata in ADTS, make MPEG-TS demuxer more resilient

This commit is contained in:
Vanilagy
2026-01-27 14:32:16 +01:00
parent cc5f7d8e8f
commit 7aa0fa465e
7 changed files with 208 additions and 59 deletions
+3 -1
View File
@@ -1,4 +1,6 @@
- Prefer functions declared using const, not using the function keyword
- Code style is tab indent with semicolons
- Mediabunny core code is contained in src/, extensions are in packages/*/, website is in docs/
- Tests: Prefer fewer, longer test files over many small ones. Test files should be named after the general catergory of thing that is being tested, not after any individual single test.
- Tests: Prefer fewer, longer test files over many small ones. Test files should be named after the general catergory of thing that is being tested, not after any individual single test.
- Avoid ifs without a {} block. So no if (cond) return;, always do if (cond) { return; }
- `type` instead of `interface` for object types
+62 -2
View File
@@ -8,9 +8,15 @@
import { aacChannelMap, aacFrequencyTable, AudioCodec } from '../codec';
import { Demuxer } from '../demuxer';
import {
ID3_V2_HEADER_SIZE,
parseId3V2Tag,
readId3V2Header,
} from '../id3';
import { Input } from '../input';
import { InputAudioTrack, InputAudioTrackBacking } from '../input-track';
import { PacketRetrievalOptions } from '../media-sink';
import { DEFAULT_TRACK_DISPOSITION, MetadataTags } from '../metadata';
import {
assert,
AsyncMutex,
@@ -20,7 +26,6 @@ import {
} from '../misc';
import { EncodedPacket, PLACEHOLDER_DATA } from '../packet';
import { readBytes, Reader } from '../reader';
import { DEFAULT_TRACK_DISPOSITION } from '../metadata';
import {
AdtsFrameHeader,
MIN_ADTS_FRAME_HEADER_SIZE,
@@ -43,6 +48,7 @@ export class AdtsDemuxer extends Demuxer {
metadataPromise: Promise<void> | null = null;
firstFrameHeader: AdtsFrameHeader | null = null;
loadedSamples: Sample[] = [];
metadataTags: MetadataTags | null = null;
tracks: InputAudioTrack[] = [];
@@ -73,6 +79,26 @@ export class AdtsDemuxer extends Demuxer {
}
async advanceReader() {
if (this.lastLoadedPos === 0) {
// Skip all ID3v2 tags at the start of the file
while (true) {
let slice = this.reader.requestSlice(this.lastLoadedPos, ID3_V2_HEADER_SIZE);
if (slice instanceof Promise) slice = await slice;
if (!slice) {
this.lastSampleLoaded = true;
return;
}
const id3V2Header = readId3V2Header(slice);
if (!id3V2Header) {
break;
}
this.lastLoadedPos = slice.filePos + id3V2Header.size;
}
}
let slice = this.reader.requestSliceRange(
this.lastLoadedPos,
MIN_ADTS_FRAME_HEADER_SIZE,
@@ -135,7 +161,41 @@ export class AdtsDemuxer extends Demuxer {
}
async getMetadataTags() {
return {}; // No tags in this one
const release = await this.readingMutex.acquire();
try {
await this.readMetadata();
if (this.metadataTags) {
return this.metadataTags;
}
this.metadataTags = {};
let currentPos = 0;
while (true) {
let headerSlice = this.reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
if (headerSlice instanceof Promise) headerSlice = await headerSlice;
if (!headerSlice) break;
const id3V2Header = readId3V2Header(headerSlice);
if (!id3V2Header) {
break;
}
let contentSlice = this.reader.requestSlice(headerSlice.filePos, id3V2Header.size);
if (contentSlice instanceof Promise) contentSlice = await contentSlice;
if (!contentSlice) break;
parseId3V2Tag(contentSlice, id3V2Header, this.metadataTags);
currentPos = headerSlice.filePos + id3V2Header.size;
}
return this.metadataTags;
} finally {
release();
}
}
}
+6 -1
View File
@@ -7,6 +7,8 @@
*/
import { parseAacAudioSpecificConfig, validateAudioChunkMetadata } from '../codec';
import { Id3V2Writer } from '../id3';
import { metadataTagsAreEmpty } from '../metadata';
import { assert, Bitstream, toUint8Array } from '../misc';
import { Muxer } from '../muxer';
import { Output, OutputAudioTrack } from '../output';
@@ -30,7 +32,10 @@ export class AdtsMuxer extends Muxer {
}
async start() {
// Nothing needed here
if (!metadataTagsAreEmpty(this.output._metadataTags)) {
const id3Writer = new Id3V2Writer(this.writer);
id3Writer.writeId3V2Tag(this.output._metadataTags);
}
}
async getMimeType() {
+19 -9
View File
@@ -266,7 +266,6 @@ export class Mp3InputFormat extends InputFormat {
if (!slice) return false;
let currentPos = 0;
let id3V2HeaderFound = false;
while (true) {
let slice = input._reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
@@ -278,7 +277,6 @@ export class Mp3InputFormat extends InputFormat {
break;
}
id3V2HeaderFound = true;
currentPos = slice.filePos + id3V2Header.size;
}
@@ -287,11 +285,6 @@ export class Mp3InputFormat extends InputFormat {
return false;
}
if (id3V2HeaderFound) {
// If there was an ID3v2 tag at the start, we can be pretty sure this is MP3 by now
return true;
}
currentPos = firstResult.startPos + firstResult.header.totalSize;
// Fine, we found one frame header, but we're still not entirely sure this is MP3. Let's check if we can find
@@ -441,8 +434,23 @@ export class FlacInputFormat extends InputFormat {
export class AdtsInputFormat extends InputFormat {
/** @internal */
async _canReadInput(input: Input) {
let currentPos = 0;
while (true) {
let slice = input._reader.requestSlice(currentPos, ID3_V2_HEADER_SIZE);
if (slice instanceof Promise) slice = await slice;
if (!slice) break;
const id3V2Header = readId3V2Header(slice);
if (!id3V2Header) {
break;
}
currentPos = slice.filePos + id3V2Header.size;
}
let slice = input._reader.requestSliceRange(
0,
currentPos,
MIN_ADTS_FRAME_HEADER_SIZE,
MAX_ADTS_FRAME_HEADER_SIZE,
);
@@ -454,8 +462,10 @@ export class AdtsInputFormat extends InputFormat {
return false;
}
currentPos += firstHeader.frameLength;
slice = input._reader.requestSliceRange(
firstHeader.frameLength,
currentPos,
MIN_ADTS_FRAME_HEADER_SIZE,
MAX_ADTS_FRAME_HEADER_SIZE,
);
+2 -1
View File
@@ -17,7 +17,7 @@
* - For Ogg files, there is no global metadata so instead, the metadata refers to the combined metadata of all tracks,
* in Vorbis-style comment headers.
* - For WAVE files, the metadata refers to the chunks within the RIFF INFO chunk.
* - For ADTS files, there is no metadata.
* - For ADTS files, the metadata refers to the ID3v2 tags.
* - For FLAC files, the metadata lives in Vorbis style in the Vorbis comment block.
* - For MPEG-TS files, metadata tags are currently not supported.
*
@@ -70,6 +70,7 @@ export type MetadataTags = {
* values. Additionally, all attached files (such as font files) are included here, where the key corresponds to
* the FileUID and the value is an {@link AttachedFile}.
* - MP3: The ID3v2 tags, or a single `'TAG'` key with the contents of the ID3v1 tag.
* - ADTS: The ID3v2 tags.
* - Ogg: The key-value string pairs from the Vorbis-style comment header (see RFC 7845, Section 5.2).
* Additionally, the `'vendor'` key refers to the vendor string within this header.
* - WAVE: The individual metadata chunks within the RIFF INFO chunk. Values are always ISO 8859-1 strings.
+60 -45
View File
@@ -63,8 +63,6 @@ import { EncodedPacket, PacketType, PLACEHOLDER_DATA } from '../packet';
import { FileSlice, readBytes, Reader, readU16Be, readU32Be, readU8 } from '../reader';
import { buildMpegTsMimeType, MpegTsStreamType, TIMESCALE, TS_PACKET_SIZE } from './mpeg-ts-misc';
const MISSING_PES_PACKET_ERROR = 'No PES packet found where one was expected.';
type ElementaryStream = {
demuxer: MpegTsDemuxer;
pid: number;
@@ -191,6 +189,26 @@ export class MpegTsDemuxer extends Demuxer {
const BYTES_BEFORE_SECTION_LENGTH = 3;
const BITS_IN_CRC_32 = 32; // Duh
// Some streams don't contain a PAT for some reason, so we must do some guesswork to figure out where
// the PMT is.
let isProbablyProgramMap = false;
if (!hasProgramMap && section.pid !== 0) {
const isPesPacket
= section.payload[0] === 0x00 && section.payload[1] === 0x00 && section.payload[2] === 0x01;
if (!isPesPacket) {
// Assume it's a PSI
const bitstream = new Bitstream(section.payload);
const pointerField = bitstream.readAlignedByte();
bitstream.skipBits(8 * pointerField);
const tableId = bitstream.readBits(8);
isProbablyProgramMap = tableId === 0x02; // 0x02 == TS_program_map_section
}
}
if (section.pid === 0 && !hasProgramAssociationTable) {
const bitstream = new Bitstream(section.payload);
const pointerField = bitstream.readAlignedByte();
@@ -220,7 +238,7 @@ export class MpegTsDemuxer extends Demuxer {
}
hasProgramAssociationTable = true;
} else if (section.pid === programMapPid && !hasProgramMap) {
} else if ((section.pid === programMapPid || isProbablyProgramMap) && !hasProgramMap) {
const bitstream = new Bitstream(section.payload);
const pointerField = bitstream.readAlignedByte();
@@ -291,6 +309,7 @@ export class MpegTsDemuxer extends Demuxer {
default: {
// If we don't recognize the codec, we don't surface the track at all. This is because
// we can't determine its metadata and also have no idea how to packetize its data.
console.warn(`Unsupported stream_type 0x${streamType.toString(16)}; ignoring stream.`);
}
}
@@ -433,10 +452,11 @@ export class MpegTsDemuxer extends Demuxer {
currentPos += this.packetStride;
}
if (!hasProgramAssociationTable) {
throw new Error('No Program Association Table found in the file.');
}
if (!hasProgramMap) {
if (!hasProgramAssociationTable) {
throw new Error('No Program Association Table found in the file.');
}
throw new Error('No Program Map Table found in the file.');
}
@@ -674,6 +694,10 @@ type PesPacket = PesPacketHeader & {
};
const readPesPacketHeader = (section: Section): PesPacketHeader | null => {
if (section.payload.byteLength < 3) {
return null;
}
const bitstream = new Bitstream(section.payload);
const startCodePrefix = bitstream.readBits(24);
@@ -994,7 +1018,9 @@ export abstract class MpegTsTrackBacking implements InputTrackBacking {
}
const pesPacketHeader = readPesPacketHeader(section);
return pesPacketHeader;
if (pesPacketHeader) {
return pesPacketHeader;
}
}
currentPos += demuxer.packetStride;
@@ -1152,17 +1178,16 @@ export abstract class MpegTsTrackBacking implements InputTrackBacking {
const section = await demuxer.readSection(currentPos, false);
if (section) {
const nextPesHeader = readPesPacketHeader(section);
if (!nextPesHeader) {
throw new Error(MISSING_PES_PACKET_ERROR);
}
if (nextPesHeader.pts > searchPts) {
break outer;
}
if (nextPesHeader) {
if (nextPesHeader.pts > searchPts) {
break outer;
}
currentPesHeader = nextPesHeader;
maybeInsertReferencePacket(this.elementaryStream, nextPesHeader);
currentPesHeader = nextPesHeader;
maybeInsertReferencePacket(this.elementaryStream, nextPesHeader);
break;
break;
}
}
}
@@ -1185,12 +1210,10 @@ export abstract class MpegTsTrackBacking implements InputTrackBacking {
const section = await demuxer.readSection(pos, false);
if (section) {
const header = readPesPacketHeader(section);
if (!header) {
throw new Error(MISSING_PES_PACKET_ERROR);
if (header) {
currentPesHeader = header;
break;
}
currentPesHeader = header;
break;
}
}
@@ -1265,13 +1288,10 @@ export abstract class MpegTsTrackBacking implements InputTrackBacking {
const section = await demuxer.readSection(currentPos, false);
if (section) {
pesHeader = readPesPacketHeader(section);
if (!pesHeader) {
throw new Error(MISSING_PES_PACKET_ERROR);
if (pesHeader) {
maybeInsertReferencePacket(this.elementaryStream, pesHeader);
break;
}
maybeInsertReferencePacket(this.elementaryStream, pesHeader);
break;
}
}
@@ -1298,12 +1318,10 @@ export abstract class MpegTsTrackBacking implements InputTrackBacking {
const section = await demuxer.readSection(pos, false);
if (section) {
const header = readPesPacketHeader(section);
if (!header) {
throw new Error(MISSING_PES_PACKET_ERROR);
if (header) {
startPesHeader = header;
break;
}
startPesHeader = header;
break;
}
}
@@ -1759,23 +1777,20 @@ class PacketReadingContext {
}
if (packetHeader.pid === this.pid) {
break;
const nextSection = await this.demuxer.readSection(currentPos, true);
if (!nextSection) {
return;
}
const nextPesPacket = readPesPacket(nextSection);
if (nextPesPacket) {
pesPacket = nextPesPacket;
break;
}
}
currentPos += this.demuxer.packetStride;
}
const nextSection = await this.demuxer.readSection(currentPos, true);
if (!nextSection) {
return;
}
const nextPesPacket = readPesPacket(nextSection);
if (!nextPesPacket) {
throw new Error(MISSING_PES_PACKET_ERROR);
}
pesPacket = nextPesPacket;
}
this.pesPackets.push(pesPacket);
+56
View File
@@ -1,6 +1,7 @@
import { expect, test } from 'vitest';
import { Output } from '../../src/output.js';
import {
AdtsOutputFormat,
FlacOutputFormat,
MkvOutputFormat,
MovOutputFormat,
@@ -478,6 +479,61 @@ test('Read and write metadata, FLAC', async () => {
expect(readTags.raw!['COMPOSER']).toBe('Hans Zimmer');
});
test('Read and write metadata, ADTS', async () => {
const originalInput = new Input({
source: new FilePathSource(path.join(__dirname, '../public/sample3.aac')),
formats: ALL_FORMATS,
});
const output = new Output({
format: new AdtsOutputFormat(),
target: new BufferTarget(),
});
const conversion = await Conversion.init({
input: originalInput,
output,
tags: {
...songMetadata,
raw: {
TXXY: 'ID3v2 goated',
},
},
});
await conversion.execute();
using input = new Input({
source: new BufferSource(output.target.buffer!),
formats: ALL_FORMATS,
});
const readTags = await input.getMetadataTags();
// ID3v2 is goated, so pretty much everything was copied:
expect(readTags.title).toBe(songMetadata.title);
expect(readTags.description).toBe(songMetadata.description);
expect(readTags.artist).toBe(songMetadata.artist);
expect(readTags.album).toBe(songMetadata.album);
expect(readTags.albumArtist).toBe(songMetadata.albumArtist);
expect(readTags.comment).toBe(songMetadata.comment);
expect(readTags.lyrics).toBe(songMetadata.lyrics);
expect(readTags.trackNumber).toBe(songMetadata.trackNumber);
expect(readTags.tracksTotal).toBe(songMetadata.tracksTotal);
expect(readTags.discNumber).toBe(songMetadata.discNumber);
expect(readTags.discsTotal).toBe(songMetadata.discsTotal);
expect(readTags.date).toEqual(readTags.date);
expect(readTags.images).toHaveLength(1);
expect(readTags.images![0]!.data).toEqual(coverArt);
expect(readTags.images![0]!.mimeType).toEqual('image/jpeg');
expect(readTags.images![0]!.kind).toEqual('coverFront');
expect(readTags.images![0]!.description).toEqual(songMetadata.images![0]!.description);
expect(readTags.images![0]!.name).toBeUndefined(); // Can't be contained in ID3v2
expect(readTags.raw!['TIT2']).toBe(songMetadata.title);
expect(readTags.raw!['APIC']).instanceOf(Uint8Array);
expect(readTags.raw!['TXXY']).toBe('ID3v2 goated');
});
test('Read and write metadata, WAVE', async () => {
const output = new Output({
format: new WavOutputFormat(),