Fix faulty MP3 bitrate and frame size calculation (fixes #134)

This commit is contained in:
Vanilagy
2025-09-19 13:20:36 +02:00
parent e88bc15b0e
commit eb1ce86187
10 changed files with 151 additions and 129 deletions
+1 -3
View File
@@ -24,9 +24,7 @@
chunked: true,
chunkSize: 2**20
});
const outputFormat = new Mediabunny.Mp4OutputFormat({
metadataFormat: 'mdta',
});
const outputFormat = new Mediabunny.Mp3OutputFormat({});
const button = document.createElement('button');
button.textContent = 'Cancel';
+12 -1
View File
@@ -14,7 +14,18 @@
source: new Mediabunny.BlobSource(file),
});
console.log(await input.getMetadataTags());
console.log(await input.computeDuration());
return;
const audioTrack = await input.getPrimaryAudioTrack();
const sink = new Mediabunny.EncodedPacketSink(audioTrack);
for await (const packet of sink.packets()) {
console.log(packet)
}
console.log("Done")
/*
+6 -6
View File
@@ -1,12 +1,12 @@
{
"name": "mediabunny",
"version": "1.17.0",
"version": "1.17.1",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "mediabunny",
"version": "1.17.0",
"version": "1.17.1",
"license": "MPL-2.0",
"workspaces": [
"packages/*"
@@ -7749,9 +7749,9 @@
}
},
"node_modules/mediabunny": {
"version": "1.16.1",
"resolved": "https://registry.npmjs.org/mediabunny/-/mediabunny-1.16.1.tgz",
"integrity": "sha512-BKA5wJaen/3g0nclyykXkS2rELsokcfOYSUfYbSZpDbChqxuITn+bCQhSTtDyMlv3nmK6KoSTo/JhcZLbF9pxQ==",
"version": "1.17.0",
"resolved": "https://registry.npmjs.org/mediabunny/-/mediabunny-1.17.0.tgz",
"integrity": "sha512-3IvrYu1Eur6IoLjygMAeVrh+5jkJxsr1RTAc7+e4QYONUt4MrUaorPIl3XtLKQdJLGcxlERrHn2KC50qBZN9mg==",
"license": "MPL-2.0",
"peer": true,
"workspaces": [
@@ -12242,7 +12242,7 @@
},
"packages/mp3-encoder": {
"name": "@mediabunny/mp3-encoder",
"version": "1.17.0",
"version": "1.17.1",
"license": "MPL-2.0",
"devDependencies": {
"@types/emscripten": "^1.40.1"
+1 -1
View File
@@ -1,7 +1,7 @@
{
"name": "mediabunny",
"author": "Vanilagy",
"version": "1.17.0",
"version": "1.17.1",
"description": "Pure TypeScript media toolkit for reading, writing, and converting media files, directly in the browser.",
"type": "module",
"workspaces": [
+1 -1
View File
@@ -1,7 +1,7 @@
{
"name": "@mediabunny/mp3-encoder",
"author": "Vanilagy",
"version": "1.17.0",
"version": "1.17.1",
"description": "MP3 encoder extension for Mediabunny, based on LAME.",
"main": "./dist/bundles/mediabunny-mp3-encoder.mjs",
"module": "./dist/bundles/mediabunny-mp3-encoder.mjs",
+3 -1
View File
@@ -28,7 +28,9 @@ class Mp3Encoder extends CustomAudioEncoder {
static override supports(codec: AudioCodec, config: AudioDecoderConfig): boolean {
return codec === 'mp3'
&& (config.numberOfChannels === 1 || config.numberOfChannels === 2)
&& Object.values(SAMPLING_RATES).some(x => x.includes(config.sampleRate));
&& Object.values(SAMPLING_RATES).some(x =>
x === config.sampleRate || (x / 2) === config.sampleRate || (x / 4) === config.sampleRate,
);
}
async init() {
+43 -43
View File
@@ -7,32 +7,20 @@
*/
export const FRAME_HEADER_SIZE = 4;
export const SAMPLING_RATES = [44100, 48000, 32000];
export const KILOBIT_RATES = [
// lowSamplingFrequency === 0
-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // layer = 0
-1, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, -1, // layer 1
-1, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384, -1, // layer = 2
-1, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, -1, // layer = 3
// 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],
};
// lowSamplingFrequency === 1
-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // layer = 0
-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1, // layer = 1
-1, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, -1, // layer = 2
-1, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, -1, // layer = 3
];
/** 'Xing' */
export const XING = 0x58696e67;
@@ -54,12 +42,21 @@ export type FrameHeader = {
audioSamplesInFrame: number;
};
export const computeMp3FrameSize = (layer: number, bitrate: number, sampleRate: number, padding: number) => {
if (layer === 3) {
// Layer 1
return Math.floor((12 * bitrate / sampleRate + padding) * 4);
} else {
return Math.floor((144 * bitrate / sampleRate) + padding);
export const computeMp3FrameSize = (
lowSamplingFrequency: number,
layer: number,
bitrate: number,
sampleRate: number,
padding: number,
) => {
if (layer === 0) {
return 0; // Not expected that this is hit
} else if (layer === 1) {
return Math.round(144 * bitrate / (sampleRate << lowSamplingFrequency)) + padding;
} else if (layer === 2) {
return Math.round(144 * bitrate / sampleRate) + padding;
} else { // layer === 3
return (Math.round(12 * bitrate / sampleRate) + padding) * 4;
}
};
@@ -93,11 +90,21 @@ export const readFrameHeader = (word: number, remainingBytes: number | null): {
return { header: null, bytesAdvanced: 1 };
}
let lowSamplingFrequency = 0;
let mpeg25 = 0;
if (secondByte & (1 << 4)) {
lowSamplingFrequency = (secondByte & (1 << 3)) ? 0 : 1;
} else {
lowSamplingFrequency = 1;
mpeg25 = 1;
}
const mpegVersionId = (secondByte >> 3) & 0x3;
const layer = (secondByte >> 1) & 0x3;
const bitrateIndex = (thirdByte >> 4) & 0xf;
const frequencyIndex = (thirdByte >> 2) & 0x3;
const frequencyIndex = ((thirdByte >> 2) & 0x3) % 3;
const padding = (thirdByte >> 1) & 0x1;
const channel = (fourthByte >> 6) & 0x3;
@@ -106,21 +113,14 @@ export const readFrameHeader = (word: number, remainingBytes: number | null): {
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) {
const kilobitRate = KILOBIT_RATES[lowSamplingFrequency * 16 * 4 + layer * 16 + bitrateIndex]!;
if (kilobitRate === -1) {
return { header: null, bytesAdvanced: 1 };
}
const bitrate = kilobitRate * 1000;
const sampleRate = SAMPLING_RATES[mpegVersionId]?.[frequencyIndex];
if (!sampleRate || sampleRate === -1) {
return { header: null, bytesAdvanced: 1 };
}
const frameLength = computeMp3FrameSize(layer, bitrate, sampleRate, padding);
const sampleRate = SAMPLING_RATES[frequencyIndex]! >> (lowSamplingFrequency + mpeg25);
const frameLength = computeMp3FrameSize(lowSamplingFrequency, layer, bitrate, sampleRate, padding);
if (remainingBytes !== null && remainingBytes < frameLength) {
// The frame doesn't fit into the rest of the file
+9 -2
View File
@@ -117,9 +117,16 @@ export class Mp3Demuxer extends Demuxer {
this.firstFrameHeader = header;
}
const sampleDuration = header.audioSamplesInFrame / header.sampleRate;
if (header.sampleRate !== this.firstFrameHeader.sampleRate) {
console.warn(
`MP3 changed sample rate mid-file: ${this.firstFrameHeader.sampleRate} Hz to ${header.sampleRate} Hz.`
+ ` Might be a bug, so please report this file.`,
);
}
const sampleDuration = header.audioSamplesInFrame / this.firstFrameHeader.sampleRate;
const sample: Sample = {
timestamp: this.nextTimestampInSamples / header.sampleRate,
timestamp: this.nextTimestampInSamples / this.firstFrameHeader.sampleRate,
duration: sampleDuration,
dataStart: result.startPos,
dataSize: header.totalSize,
+1
View File
@@ -84,6 +84,7 @@ export class Mp3Muxer extends Muxer {
mpegVersionId: header.mpegVersionId,
layer: header.layer,
frequencyIndex: header.frequencyIndex,
sampleRate: header.sampleRate,
channel: header.channel,
modeExtension: header.modeExtension,
copyright: header.copyright,
+74 -71
View File
@@ -12,9 +12,7 @@ import {
computeMp3FrameSize,
encodeSynchsafe,
getXingOffset,
MPEG_V1_BITRATES,
MPEG_V2_BITRATES,
SAMPLING_RATES,
KILOBIT_RATES,
XING,
} from '../../shared/mp3-misc';
import { Id3V2TextEncoding } from './mp3-reader';
@@ -23,6 +21,7 @@ export type XingFrameData = {
mpegVersionId: number;
layer: number;
frequencyIndex: number;
sampleRate: number;
channel: number;
modeExtension: number;
copyright: number;
@@ -62,6 +61,78 @@ export class Mp3Writer {
this.writer.write(this.helper.subarray(0, text.length));
}
writeXingFrame(data: XingFrameData) {
const startPos = this.writer.getPos();
const firstByte = 0xff;
const secondByte = 0xe0 | (data.mpegVersionId << 3) | (data.layer << 1);
const lowSamplingFrequency = data.mpegVersionId & 1;
const padding = 0;
const neededBytes = 155;
let bitrateIndex = -1;
const bitrateOffset = lowSamplingFrequency * 16 * 4 + data.layer * 16;
// Let's find the lowest bitrate for which the frame size is sufficiently large to fit all the data
for (let i = 0; i < 16; i++) {
const kbr = KILOBIT_RATES[bitrateOffset + i]!;
const size = computeMp3FrameSize(lowSamplingFrequency, data.layer, 1000 * kbr, data.sampleRate, padding);
if (size >= neededBytes) {
bitrateIndex = i;
break;
}
}
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 kilobitRate = KILOBIT_RATES[bitrateOffset + bitrateIndex]!;
const frameSize = computeMp3FrameSize(
lowSamplingFrequency, data.layer, 1000 * kilobitRate, data.sampleRate, padding,
);
this.writer.seek(startPos + frameSize);
}
writeSynchsafeU32(value: number) {
this.writeU32(encodeSynchsafe(value));
}
@@ -170,72 +241,4 @@ export class Mp3Writer {
this.writer.write(imageData);
}
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(data.layer, 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(data.layer, 1000 * bitrates[bitrateIndex]!, sampleRate, padding);
this.writer.seek(startPos + frameSize);
}
}