Implement MP3 muxer

This commit is contained in:
Vanilagy
2025-01-26 11:45:56 +01:00
parent 7205d9445b
commit d3566305df
11 changed files with 482 additions and 148 deletions
+34
View File
@@ -16,11 +16,45 @@
source
});
const output = new Metamuxer.Output({
format: new Metamuxer.Mp3OutputFormat(),
target: new Metamuxer.BufferTarget()
});
const audioTrack = await input.getPrimaryAudioTrack();
const sink = new Metamuxer.EncodedAudioSampleSink(audioTrack);
const mediaSource = new Metamuxer.EncodedAudioSampleSource(await audioTrack.getCodec());
output.addAudioTrack(mediaSource);
output.start();
for await (const sample of sink.samples()) {
await mediaSource.digest(sample);
}
await output.finalize();
function download(blob, filename) {
const url = URL.createObjectURL(blob);
const a = document.createElement('a');
a.href = url;
a.download = filename;
a.click();
URL.revokeObjectURL(url);
}
console.log(output.target.buffer)
download(new Blob([output.target.buffer]), 'converted.mp3');
/*
const videoTrack = await input.getPrimaryVideoTrack();
const sink = new Metamuxer.EncodedVideoSampleSink(videoTrack);
console.log(await sink.getSample(16));
console.log(await sink.getFirstSample());
*/
/*
const source = new Metamuxer.BufferSource(await file.arrayBuffer()) ?? new Metamuxer.BlobSource(file);
+1
View File
@@ -18,6 +18,7 @@ export {
MkvOutputFormatOptions,
WebMOutputFormat,
WebMOutputFormatOptions,
Mp3OutputFormat,
WaveOutputFormat,
TrackCountLimits,
InclusiveRange,
+33 -33
View File
@@ -179,39 +179,6 @@ export class WebMInputFormat extends MatroskaInputFormat {
}
}
/** @public */
export class WaveInputFormat extends InputFormat {
async _canReadInput(input: Input) {
const sourceSize = await input._mainReader.source._getSize();
if (sourceSize < 12) {
return false;
}
const riffReader = new RiffReader(input._mainReader);
const riffType = riffReader.readAscii(4);
if (riffType !== 'RIFF' && riffType !== 'RIFX') {
return false;
}
riffReader.pos = 8;
const format = riffReader.readAscii(4);
return format === 'WAVE';
}
/** @internal */
_createDemuxer(input: Input) {
return new WaveDemuxer(input);
}
getName() {
return 'WAVE';
}
getMimeType() {
return 'audio/wav';
}
}
/** @public */
export class Mp3InputFormat extends InputFormat {
async _canReadInput(input: Input) {
@@ -248,6 +215,39 @@ export class Mp3InputFormat extends InputFormat {
}
}
/** @public */
export class WaveInputFormat extends InputFormat {
async _canReadInput(input: Input) {
const sourceSize = await input._mainReader.source._getSize();
if (sourceSize < 12) {
return false;
}
const riffReader = new RiffReader(input._mainReader);
const riffType = riffReader.readAscii(4);
if (riffType !== 'RIFF' && riffType !== 'RIFX') {
return false;
}
riffReader.pos = 8;
const format = riffReader.readAscii(4);
return format === 'WAVE';
}
/** @internal */
_createDemuxer(input: Input) {
return new WaveDemuxer(input);
}
getName() {
return 'WAVE';
}
getMimeType() {
return 'audio/wav';
}
}
/** @public */
export const MP4 = new Mp4InputFormat();
/** @public */
+6 -8
View File
@@ -5,7 +5,8 @@ import { InputAudioTrack, InputAudioTrackBacking } from '../input-track';
import { SampleRetrievalOptions } from '../media-sink';
import { assert, binarySearchExact, binarySearchLessOrEqual, last, UNDETERMINED_LANGUAGE } from '../misc';
import { EncodedAudioSample, PLACEHOLDER_DATA } from '../sample';
import { FrameHeader, Mp3Reader } from './mp3-reader';
import { FrameHeader, getXingOffset, INFO, XING } from './mp3-misc';
import { Mp3Reader } from './mp3-reader';
const AUDIO_SAMPLES_PER_FRAME = 1152;
@@ -54,13 +55,10 @@ export class Mp3Demuxer extends Demuxer {
break;
}
const xingOffset = header.mpegVersionId === 3
? (header.channelCount === 1 ? 21 : 36)
: (header.channelCount === 1 ? 13 : 21);
const xingOffset = getXingOffset(header.mpegVersionId, header.channel);
this.reader.pos = header.startPos + xingOffset;
const word = this.reader.readU32();
const isXing = word === 0x58696e67 // 'Xing'
|| word === 0x496e666f; // 'Info'
const isXing = word === XING || word === INFO;
this.reader.pos = header.startPos + header.totalSize - 1; // -1 in case the frame is 1 byte too short
@@ -137,7 +135,7 @@ class Mp3AudioTrackBacking implements InputAudioTrackBacking {
async getNumberOfChannels() {
assert(this.demuxer.firstFrameHeader);
return this.demuxer.firstFrameHeader.channelCount;
return this.demuxer.firstFrameHeader.channel === 3 ? 1 : 2;
}
async getSampleRate() {
@@ -150,7 +148,7 @@ class Mp3AudioTrackBacking implements InputAudioTrackBacking {
return {
codec: 'mp3',
numberOfChannels: this.demuxer.firstFrameHeader.channelCount,
numberOfChannels: this.demuxer.firstFrameHeader.channel === 3 ? 1 : 2,
sampleRate: this.demuxer.firstFrameHeader.sampleRate,
};
}
+130
View File
@@ -0,0 +1,130 @@
export const FRAME_HEADER_SIZE = 4;
// These are in kbps:
export const MPEG_V1_BITRATES: Record<number, number[]> = {
// Layer 3
1: [-1, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, -1],
// Layer 2
2: [-1, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384, -1],
// Layer 1
3: [-1, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, -1],
};
export const MPEG_V2_BITRATES: Record<number, number[]> = {
// Layer 3
1: [-1, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, -1],
// Layer 2
2: [-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1],
// Layer 1
3: [-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1],
};
export const SAMPLING_RATES: Record<number, number[]> = {
// MPEG Version 2.5
0: [11025, 12000, 8000, -1],
// MPEG Version 2 (ISO/IEC 13818-3)
2: [22050, 24000, 16000, -1],
// MPEG Version 1 (ISO/IEC 11172-3)
3: [44100, 48000, 32000, -1],
};
/** 'Xing' */
export const XING = 0x58696e67;
/** 'Info' */
export const INFO = 0x496e666f;
export type FrameHeader = {
startPos: number;
mpegVersionId: number;
layer: number;
bitrate: number;
frequencyIndex: number;
sampleRate: number;
channel: number;
modeExtension: number;
copyright: number;
original: number;
emphasis: number;
totalSize: number;
dataStart: number;
dataSize: number;
};
export const computeMp3FrameSize = (bitrate: number, sampleRate: number, padding: number) => {
return Math.floor((144 * bitrate / sampleRate) + padding);
};
export const getXingOffset = (mpegVersionId: number, channel: number) => {
return mpegVersionId === 3
? (channel === 3 ? 21 : 36)
: (channel === 3 ? 13 : 21);
};
export const readFrameHeader = (word: number, reader: { pos: number; fileSize: number | null }): FrameHeader | null => {
const startPos = reader.pos;
const firstByte = word >>> 24;
const secondByte = (word >>> 16) & 0xff;
const thirdByte = (word >>> 8) & 0xff;
const fourthByte = word & 0xff;
if (firstByte !== 0xff && secondByte !== 0xff && thirdByte !== 0xff && fourthByte !== 0xff) {
reader.pos += 4;
return null;
}
reader.pos += 1;
if ((secondByte & 0xe0) !== 0xe0) {
return null;
}
const mpegVersionId = (secondByte >> 3) & 0x3;
const layer = (secondByte >> 1) & 0x3;
const bitrateIndex = (thirdByte >> 4) & 0xf;
const frequencyIndex = (thirdByte >> 2) & 0x3;
const padding = (thirdByte >> 1) & 0x1;
const channel = (fourthByte >> 6) & 0x3;
const modeExtension = (fourthByte >> 4) & 0x3;
const copyright = (fourthByte >> 3) & 0x1;
const original = (fourthByte >> 2) & 0x1;
const emphasis = fourthByte & 0x3;
const kilobitRate = mpegVersionId === 3
? MPEG_V1_BITRATES[layer]?.[bitrateIndex]
: MPEG_V2_BITRATES[layer]?.[bitrateIndex];
if (!kilobitRate || kilobitRate === -1) {
return null;
}
const bitrate = kilobitRate * 1000;
const sampleRate = SAMPLING_RATES[mpegVersionId]?.[frequencyIndex];
if (!sampleRate || sampleRate === -1) {
return null;
}
const frameLength = computeMp3FrameSize(bitrate, sampleRate, padding);
if (reader.fileSize !== null && reader.fileSize - startPos < frameLength) {
// The frame doesn't fit into the rest of the file
return null;
}
return {
startPos: startPos,
mpegVersionId,
layer,
bitrate,
frequencyIndex,
sampleRate,
channel,
modeExtension,
copyright,
original,
emphasis,
totalSize: frameLength,
dataStart: startPos + 4,
dataSize: frameLength - 4,
};
};
+125
View File
@@ -0,0 +1,125 @@
import { assert } from '../misc';
import { Muxer } from '../muxer';
import { Output, OutputAudioTrack } from '../output';
import { EncodedAudioSample } from '../sample';
import { Writer } from '../writer';
import { getXingOffset, INFO, readFrameHeader, XING } from './mp3-misc';
import { Mp3Writer, XingFrameData } from './mp3-writer';
export class Mp3Muxer extends Muxer {
private writer: Writer;
private mp3Writer: Mp3Writer;
private xingFrameData: XingFrameData | null = null;
private frameCount = 0;
private framePositions: number[] = [];
constructor(output: Output) {
super(output);
this.writer = output._writer;
this.mp3Writer = new Mp3Writer(output._writer);
}
async start() {
// Nothing needed here
}
async addEncodedVideoSample() {
throw new Error('MP3 does not support video.');
}
async addEncodedAudioSample(
track: OutputAudioTrack,
sample: EncodedAudioSample,
) {
const release = await this.mutex.acquire();
try {
if (!this.xingFrameData) {
const view = new DataView(sample.data.buffer, sample.data.byteOffset, sample.data.byteLength);
if (view.byteLength < 4) {
throw new Error('Invalid MP3 header in sample.');
}
const word = view.getUint32(0, false);
const header = readFrameHeader(word, { pos: 0, fileSize: null });
if (!header) {
throw new Error('Invalid MP3 header in sample.');
}
const xingOffset = getXingOffset(header.mpegVersionId, header.channel);
if (view.byteLength >= xingOffset + 4) {
const word = view.getUint32(xingOffset, false);
const isXing = word === XING || word === INFO;
if (isXing) {
// This is not a data frame, so let's completely ignore this sample
return;
}
}
this.xingFrameData = {
mpegVersionId: header.mpegVersionId,
layer: header.layer,
frequencyIndex: header.frequencyIndex,
channel: header.channel,
modeExtension: header.modeExtension,
copyright: header.copyright,
original: header.original,
emphasis: header.emphasis,
frameCount: null,
fileSize: null,
toc: null,
};
// Write a Xing frame because this muxer doesn't make any bitrate constraints, meaning we don't know if
// this will be a constant or variable bitrate file. Therefore, always write the Xing frame.
this.mp3Writer.writeXingFrame(this.xingFrameData);
this.frameCount++;
}
this.validateAndNormalizeTimestamp(track, sample.timestamp, sample.type === 'key');
this.framePositions.push(this.writer.getPos());
this.writer.write(sample.data);
this.frameCount++;
await this.writer.flush();
} finally {
release();
}
}
async addSubtitleCue() {
throw new Error('MP3 does not support subtitles.');
}
async finalize() {
if (!this.xingFrameData) {
return;
}
const endPos = this.writer.getPos();
this.writer.seek(0);
const toc = new Uint8Array(100);
for (let i = 0; i < 100; i++) {
const index = Math.floor(this.framePositions.length * (i / 100));
assert(index !== -1 && index < this.framePositions.length);
const byteOffset = this.framePositions[index]!;
toc[i] = 256 * (byteOffset / endPos);
}
this.xingFrameData.frameCount = this.frameCount;
this.xingFrameData.fileSize = endPos;
this.xingFrameData.toc = toc;
this.mp3Writer.writeXingFrame(this.xingFrameData);
this.writer.seek(endPos);
}
}
+5 -96
View File
@@ -1,44 +1,6 @@
import { assert } from '../misc';
import { Reader } from '../reader';
const FRAME_HEADER_SIZE = 4;
// These are in kbps:
const MPEG_V1_BITRATES: Record<number, number[]> = {
// Layer 3
1: [-1, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, -1],
// Layer 2
2: [-1, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384, -1],
// Layer 1
3: [-1, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, -1],
};
const MPEG_V2_BITRATES: Record<number, number[]> = {
// Layer 3
1: [-1, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, -1],
// Layer 2
2: [-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1],
// Layer 1
3: [-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1],
};
const SAMPLING_RATES: Record<number, number[]> = {
// MPEG Version 2.5
0: [11025, 12000, 8000, -1],
// MPEG Version 2 (ISO/IEC 13818-3)
2: [22050, 24000, 16000, -1],
// MPEG Version 1 (ISO/IEC 11172-3)
3: [44100, 48000, 32000, -1],
};
export type FrameHeader = {
startPos: number;
mpegVersionId: number;
bitrate: number;
sampleRate: number;
channelCount: number;
totalSize: number;
dataStart: number;
dataSize: number;
};
import { FRAME_HEADER_SIZE, FrameHeader, readFrameHeader } from './mp3-misc';
export class Mp3Reader {
pos = 0;
@@ -96,66 +58,13 @@ export class Mp3Reader {
until ??= this.fileSize;
while (this.pos < until - FRAME_HEADER_SIZE) {
const startPos = this.pos;
const word = this.readU32();
this.pos -= 4;
const firstByte = word >>> 24;
const secondByte = (word >>> 16) & 0xff;
const thirdByte = (word >>> 8) & 0xff;
const fourthByte = word & 0xff;
if (firstByte !== 0xff && secondByte !== 0xff && thirdByte !== 0xff && fourthByte !== 0xff) {
continue;
const header = readFrameHeader(word, this);
if (header) {
return header;
}
this.pos -= 3; // For when we continue
if ((secondByte & 0xe0) !== 0xe0) {
continue;
}
const mpegVersionId = (secondByte >> 3) & 0x3;
const layer = (secondByte >> 1) & 0x3;
const bitrateIndex = (thirdByte >> 4) & 0xf;
const frequencyIndex = (thirdByte >> 2) & 0x3;
const padding = (thirdByte >> 1) & 0x1;
const channel = (fourthByte >> 6) & 0x3;
const kilobitRate = mpegVersionId === 3
? MPEG_V1_BITRATES[layer]?.[bitrateIndex]
: MPEG_V2_BITRATES[layer]?.[bitrateIndex];
if (!kilobitRate || kilobitRate === -1) {
continue;
}
const bitrate = kilobitRate * 1000;
const sampleRate = SAMPLING_RATES[mpegVersionId]?.[frequencyIndex];
if (!sampleRate || sampleRate === -1) {
continue;
}
const channelCount = channel === 3 ? 1 : 2;
const frameLength = Math.floor((144 * bitrate / sampleRate) + padding);
if (this.fileSize - startPos < frameLength) {
// The frame doesn't fit into the rest of the file
return null;
}
return {
startPos,
mpegVersionId,
bitrate,
sampleRate,
channelCount,
totalSize: frameLength,
dataStart: startPos + 4,
dataSize: frameLength - 4,
};
}
return null;
+104
View File
@@ -0,0 +1,104 @@
import { Writer } from '../writer';
import {
computeMp3FrameSize,
getXingOffset,
MPEG_V1_BITRATES,
MPEG_V2_BITRATES,
SAMPLING_RATES,
XING,
} from './mp3-misc';
export type XingFrameData = {
mpegVersionId: number;
layer: number;
frequencyIndex: number;
channel: number;
modeExtension: number;
copyright: number;
original: number;
emphasis: number;
frameCount: number | null;
fileSize: number | null;
toc: Uint8Array | null;
};
export class Mp3Writer {
private helper = new Uint8Array(8);
private helperView = new DataView(this.helper.buffer);
constructor(private writer: Writer) {}
writeU32(value: number) {
this.helperView.setUint32(0, value, false);
this.writer.write(this.helper.subarray(0, 4));
}
writeXingFrame(data: XingFrameData) {
const startPos = this.writer.getPos();
const firstByte = 0xff;
const secondByte = 0xe0 | (data.mpegVersionId << 3) | (data.layer << 1);
const bitrateGroup = data.mpegVersionId === 3 ? MPEG_V1_BITRATES : MPEG_V2_BITRATES;
const bitrates = bitrateGroup?.[data.layer];
if (!bitrates) {
throw new Error('Invalid MPEG version and layer combination.');
}
const sampleRate = SAMPLING_RATES[data.mpegVersionId]?.[data.frequencyIndex];
if (!sampleRate || sampleRate === -1) {
throw new Error('Invalid MPEG version and frequency index combination.');
}
const padding = 0;
const neededBytes = 155;
// Let's find the lowest bitrate for which the frame size is sufficiently large to fit all the data
const bitrateIndex = bitrates.findIndex((kbr) => {
return computeMp3FrameSize(1000 * kbr, sampleRate, padding) >= neededBytes;
});
if (bitrateIndex === -1) {
throw new Error('No suitable bitrate found.');
}
const thirdByte = (bitrateIndex << 4) | (data.frequencyIndex << 2) | padding << 1;
const fourthByte = (data.channel << 6)
| (data.modeExtension << 4)
| (data.copyright << 3)
| (data.original << 2)
| data.emphasis;
this.helper[0] = firstByte;
this.helper[1] = secondByte;
this.helper[2] = thirdByte;
this.helper[3] = fourthByte;
this.writer.write(this.helper.subarray(0, 4));
const xingOffset = getXingOffset(data.mpegVersionId, data.channel);
this.writer.seek(startPos + xingOffset);
this.writeU32(XING);
let flags = 0;
if (data.frameCount !== null) {
flags |= 1;
}
if (data.fileSize !== null) {
flags |= 2;
}
if (data.toc !== null) {
flags |= 4;
}
this.writeU32(flags);
this.writeU32(data.frameCount ?? 0);
this.writeU32(data.fileSize ?? 0);
this.writer.write(data.toc ?? new Uint8Array(100));
const frameSize = computeMp3FrameSize(1000 * bitrates[bitrateIndex]!, sampleRate, padding);
this.writer.seek(startPos + frameSize);
}
}
+2 -2
View File
@@ -14,12 +14,12 @@ export abstract class Muxer {
abstract start(): Promise<void>;
abstract addEncodedVideoSample(
track: OutputVideoTrack,
chunk: EncodedVideoSample,
sample: EncodedVideoSample,
meta?: EncodedVideoChunkMetadata
): Promise<void>;
abstract addEncodedAudioSample(
track: OutputAudioTrack,
chunk: EncodedAudioSample,
sample: EncodedAudioSample,
meta?: EncodedAudioChunkMetadata
): Promise<void>;
abstract addSubtitleCue(track: OutputSubtitleTrack, cue: SubtitleCue, meta?: SubtitleMetadata): Promise<void>;
+35
View File
@@ -11,6 +11,7 @@ import {
} from './codec';
import { IsobmffMuxer } from './isobmff/isobmff-muxer';
import { MatroskaMuxer } from './matroska/matroska-muxer';
import { Mp3Muxer } from './mp3/mp3-muxer';
import { Muxer } from './muxer';
import { Output, TrackType } from './output';
import { WaveMuxer } from './wave/wave-muxer';
@@ -256,6 +257,40 @@ export class WebMOutputFormat extends MkvOutputFormat {
}
}
/** @public */
export class Mp3OutputFormat extends OutputFormat {
/** @internal */
_createMuxer(output: Output) {
return new Mp3Muxer(output);
}
/** @internal */
_getName() {
return 'MP3';
}
getSupportedTrackCounts(): TrackCountLimits {
return {
video: { min: 0, max: 0 },
audio: { min: 1, max: 1 },
subtitle: { min: 0, max: 0 },
total: { min: 1, max: 1 },
};
}
getFileExtension() {
return '.mp3';
}
getSupportedCodecs() {
return Mp3OutputFormat.getSupportedCodecs();
}
static getSupportedCodecs(): MediaCodec[] {
return ['mp3'];
}
}
/** @public */
export class WaveOutputFormat extends OutputFormat {
/** @internal */
+7 -9
View File
@@ -24,7 +24,7 @@ export class WaveMuxer extends Muxer {
}
async addEncodedVideoSample() {
throw new Error('WAVE format does not support video.');
throw new Error('WAVE does not support video.');
}
async addEncodedAudioSample(
@@ -46,19 +46,17 @@ export class WaveMuxer extends Muxer {
this.validateAndNormalizeTimestamp(track, sample.timestamp, sample.type === 'key');
if (sample.data) {
this.output._writer.write(sample.data);
this.dataSize += sample.data.byteLength;
}
this.writer.write(sample.data);
this.dataSize += sample.data.byteLength;
await this.output._writer.flush();
await this.writer.flush();
} finally {
release();
}
}
async addSubtitleCue() {
throw new Error('WAVE format does not support subtitles.');
throw new Error('WAVE does not support subtitles.');
}
private writeHeader(track: OutputAudioTrack, config: AudioDecoderConfig) {
@@ -102,7 +100,7 @@ export class WaveMuxer extends Muxer {
}
async finalize() {
const currentPos = this.writer.getPos();
const endPos = this.writer.getPos();
// Write file size
this.writer.seek(4);
@@ -112,6 +110,6 @@ export class WaveMuxer extends Muxer {
this.writer.seek(40);
this.riffWriter.writeU32(this.dataSize);
this.writer.seek(currentPos);
this.writer.seek(endPos);
}
}