Skip to content

Commit

Permalink
Refactor Track by moving nested functions to instance functions
Browse files Browse the repository at this point in the history
  • Loading branch information
@mikeyaworski
mikeyaworski committed Jun 27, 2024
1 parent 323c295 commit d17ac76
Showing 1 changed file with 149 additions and 149 deletions.
298 changes: 149 additions & 149 deletions src/commands/player/track.ts
View file
Original file line number Diff line number Diff line change
Expand Up @@ -83,193 +83,193 @@ export default class Track {
}
}

public async createAudioResource(options: AudioResourceOptions): Promise<AudioResource<Track>> {
const { seek, speed, shouldNormalizeAudio = true } = options;
private encodeStream(stream: Readable | NodeJS.ReadableStream, options: AudioResourceOptions): AudioResource<Track> {
const { speed, shouldNormalizeAudio = true } = options;

const encodeStream = (stream: Readable | NodeJS.ReadableStream): AudioResource<Track> => {
// https://ffmpeg.org/ffmpeg-filters.html#loudnorm
// https://k.ylo.ph/2016/04/04/loudnorm.html
const loudNormOptions = [
['I', '-30.0'], // Set integrated loudness target. Range is -70.0 - -5.0. Default value is -24.0.
['LRA', '7.0'], // Set loudness range target. Range is 1.0 - 50.0. Default value is 7.0.
['TP', '-2.0'], // Set maximum true peak. Range is -9.0 - +0.0. Default value is -2.0.
].map(([key, value]) => `${key}=${value}`);
// https://ffmpeg.org/ffmpeg-filters.html#loudnorm
// https://k.ylo.ph/2016/04/04/loudnorm.html
const loudNormOptions = [
['I', '-30.0'], // Set integrated loudness target. Range is -70.0 - -5.0. Default value is -24.0.
['LRA', '7.0'], // Set loudness range target. Range is 1.0 - 50.0. Default value is 7.0.
['TP', '-2.0'], // Set maximum true peak. Range is -9.0 - +0.0. Default value is -2.0.
].map(([key, value]) => `${key}=${value}`);

const audioFilters: [string, string][] = filterOutFalsy([
this.variant !== TrackVariant.TEXT && Boolean(speed) && ['atempo', String(speed)],
this.variant !== TrackVariant.TEXT && shouldNormalizeAudio && ['loudnorm', `${loudNormOptions.join(':')}`],
]);
const audioFilters: [string, string][] = filterOutFalsy([
this.variant !== TrackVariant.TEXT && Boolean(speed) && ['atempo', String(speed)],
this.variant !== TrackVariant.TEXT && shouldNormalizeAudio && ['loudnorm', `${loudNormOptions.join(':')}`],
]);

const audioFilterArg: [string, string] | null = audioFilters.length
? ['-filter:a', audioFilters.map(([name, value]) => `${name}=${value}`).join(',')]
: null;
const audioFilterArg: [string, string] | null = audioFilters.length
? ['-filter:a', audioFilters.map(([name, value]) => `${name}=${value}`).join(',')]
: null;

const ffmpegArgs: ([string, string] | [string] | false | null)[] = [
['-analyzeduration', '0'],
['-loglevel', '0'],
['-f', 's16le'],
['-ar', '48000'],
['-ac', '2'],
audioFilterArg,
];
const transcoder = new prism.FFmpeg({
args: filterOutFalsy(ffmpegArgs.flat()),
});
const s16le = stream.pipe(transcoder);
const encoder = new prism.opus.Encoder({ rate: 48000, channels: 2, frameSize: 960 });
const opus = s16le.pipe(encoder);
const resource = createAudioResource(opus, { metadata: this, inputType: StreamType.Opus });
resource.playStream.on('close', () => {
// Clean up processes to avoid memory leaks
transcoder.destroy();
});
return resource;
};
const ffmpegArgs: ([string, string] | [string] | false | null)[] = [
['-analyzeduration', '0'],
['-loglevel', '0'],
['-f', 's16le'],
['-ar', '48000'],
['-ac', '2'],
audioFilterArg,
];
const transcoder = new prism.FFmpeg({
args: filterOutFalsy(ffmpegArgs.flat()),
});
const s16le = stream.pipe(transcoder);
const encoder = new prism.opus.Encoder({ rate: 48000, channels: 2, frameSize: 960 });
const opus = s16le.pipe(encoder);
const resource = createAudioResource(opus, { metadata: this, inputType: StreamType.Opus });
resource.playStream.on('close', () => {
// Clean up processes to avoid memory leaks
transcoder.destroy();
});
return resource;
}

private async tryPlayDl(options: AudioResourceOptions): Promise<AudioResource<Track>> {
// Cookies for play-dl are set in .data/youtube.data
const tryPlayDl = async (): Promise<AudioResource<Track>> => {
const source = options.seek
? await play.stream(this.value, {
seek: options.seek,
})
: await play.stream(this.value);
const audioResource = await createAudioResource(source.stream, {
metadata: this,
inputType: source.type,
});
return audioResource;
};
const source = options.seek
? await play.stream(this.value, {
seek: options.seek,
})
: await play.stream(this.value);
const audioResource = await createAudioResource(source.stream, {
metadata: this,
inputType: source.type,
});
return audioResource;
}

const tryYtdlExec = async (): Promise<AudioResource<Track>> => {
// https://github.com/discordjs/voice/blob/f1869a9af5a44ec9a4f52c2dd282352b1521427d/examples/music-bot/src/music/track.ts#L46-L76
return new Promise((resolve, reject) => {
const process = ytdlExec.exec(this.value, {
output: '-',
quiet: true,
format: 'bestaudio[ext=webm][acodec=opus][asr=48000]/bestaudio',
cookies: './.data/cookies.txt',
// @ts-expect-error The library has incomplete typing for flags
forceKeyframesAtCuts: true,
downloadSections: seek ? `*${seek}-inf` : '*from-url',
}, {
// Pipe stdout and stderr to the parent process. Ignore stdin.
// Obviously stdout is for the audio data, and stderr is for any error output.
stdio: ['ignore', 'pipe', 'pipe'],
// Detached so any child processes created by this yt-dlp process (e.g. FFmpeg for YouTube livestreams)
// will go into a process group that we can cleanup.
detached: true,
});
if (!process.stdout) {
reject(new Error('No stdout'));
private async tryYtdlExec(options: AudioResourceOptions): Promise<AudioResource<Track>> {
// https://github.com/discordjs/voice/blob/f1869a9af5a44ec9a4f52c2dd282352b1521427d/examples/music-bot/src/music/track.ts#L46-L76
return new Promise((resolve, reject) => {
const process = ytdlExec.exec(this.value, {
output: '-',
quiet: true,
format: 'bestaudio[ext=webm][acodec=opus][asr=48000]/bestaudio',
cookies: './.data/cookies.txt',
// @ts-expect-error The library has incomplete typing for flags
forceKeyframesAtCuts: true,
downloadSections: options.seek ? `*${options.seek}-inf` : '*from-url',
}, {
// Pipe stdout and stderr to the parent process. Ignore stdin.
// Obviously stdout is for the audio data, and stderr is for any error output.
stdio: ['ignore', 'pipe', 'pipe'],
// Detached so any child processes created by this yt-dlp process (e.g. FFmpeg for YouTube livestreams)
// will go into a process group that we can cleanup.
detached: true,
});
if (!process.stdout) {
reject(new Error('No stdout'));
return;
}
const killProcesses = () => {
const childProcessesGroupId = process.pid;
if (!process.killed) process.kill();
try {
// If --download-sections flag is used (seeking), we must use SIGKILL (instead of SIGTERM) to kill it immediately
if (childProcessesGroupId != null) killNodeProcess(-childProcessesGroupId, 'SIGKILL');
} catch (err) {
error(err);
}
};
const stream = process.stdout;
const onError = (err: unknown) => {
// When there is a null or 0 exit code, it means that the process was terminated intentionally and without error
if (typeof err === 'object' && err && 'exitCode' in err && !err.exitCode) {
return;
}
const killProcesses = () => {
const childProcessesGroupId = process.pid;
if (!process.killed) process.kill();
try {
// If --download-sections flag is used (seeking), we must use SIGKILL (instead of SIGTERM) to kill it immediately
if (childProcessesGroupId != null) killNodeProcess(-childProcessesGroupId, 'SIGKILL');
} catch (err) {
error(err);
}
};
const stream = process.stdout;
const onError = (err: unknown) => {
// When there is a null or 0 exit code, it means that the process was terminated intentionally and without error
if (typeof err === 'object' && err && 'exitCode' in err && !err.exitCode) {
return;
}
killProcesses();
stream.resume();
reject(err);
error(err);
};
killProcesses();
stream.resume();
reject(err);
error(err);
};

process.catch(onError);
process.on('error', onError); // This may be redundant, but it doesn't seem to ever get called
process.catch(onError);
process.on('error', onError); // This may be redundant, but it doesn't seem to ever get called

process.once('spawn', () => {
// What we are doing:
// 1. Using yt-dlp to get a raw stream of data from YouTube
// 2. Feeding the stream into an FFmpeg transcoder (transcoding the YouTube audio stream to a s16le format for the raw audio)
// 3. Manipulating the stream in various ways with FFmpeg (e.g. seeking, playback speed, loudness normalization)
// 4. Feeding the raw audio data (s16le) into an Opus encoder with the correct settings for Discord's API
// 5. Creating an audio resource with that Opus encoder
const resource = encodeStream(stream);
resource.playStream.on('close', () => {
// Clean up processes to avoid memory leaks
killProcesses();
});
return resolve(resource);
process.once('spawn', () => {
// What we are doing:
// 1. Using yt-dlp to get a raw stream of data from YouTube
// 2. Feeding the stream into an FFmpeg transcoder (transcoding the YouTube audio stream to a s16le format for the raw audio)
// 3. Manipulating the stream in various ways with FFmpeg (e.g. seeking, playback speed, loudness normalization)
// 4. Feeding the raw audio data (s16le) into an Opus encoder with the correct settings for Discord's API
// 5. Creating an audio resource with that Opus encoder
const resource = this.encodeStream(stream, options);
resource.playStream.on('close', () => {
// Clean up processes to avoid memory leaks
killProcesses();
});
return resolve(resource);
});
};
});
}

const openAiTextToSpeech: () => Promise<AudioResource<Track>> = async () => {
if (!openai) throw new Error('OpenAI not configured');
const response = await openai.audio.speech.create({
model: 'tts-1',
voice: 'alloy',
input: this.value,
response_format: 'opus',
speed: speed || 1.0,
});
if (!response || !response.body) {
throw new Error('Could not get speech');
}
const stream = response.body;
return encodeStream(stream);
};
private async openAiTextToSpeech(options: AudioResourceOptions): Promise<AudioResource<Track>> {
if (!openai) throw new Error('OpenAI not configured');
const response = await openai.audio.speech.create({
model: 'tts-1',
voice: 'alloy',
input: this.value,
response_format: 'opus',
speed: options.speed || 1.0,
});
if (!response || !response.body) {
throw new Error('Could not get speech');
}
const stream = response.body;
return this.encodeStream(stream, options);
}

const googleTextToSpeech: () => Promise<AudioResource<Track>> = async () => {
const client = new GoogleTextToSpeechClient();
const [response] = await client.synthesizeSpeech({
input: {
text: this.value,
},
voice: {
languageCode: 'en-US',
ssmlGender: 'NEUTRAL',
},
audioConfig: {
audioEncoding: 'OGG_OPUS',
},
});
if (!response.audioContent || typeof response.audioContent === 'string') {
error(`Failed to convert text to speech: ${response.audioContent}`);
throw new Error('Failed to convert text to speech.');
}
const stream = new PassThrough();
stream.end(Buffer.from(response.audioContent));
return encodeStream(stream);
};
private async googleTextToSpeech(options: AudioResourceOptions): Promise<AudioResource<Track>> {
const client = new GoogleTextToSpeechClient();
const [response] = await client.synthesizeSpeech({
input: {
text: this.value,
},
voice: {
languageCode: 'en-US',
ssmlGender: 'NEUTRAL',
},
audioConfig: {
audioEncoding: 'OGG_OPUS',
},
});
if (!response.audioContent || typeof response.audioContent === 'string') {
error(`Failed to convert text to speech: ${response.audioContent}`);
throw new Error('Failed to convert text to speech.');
}
const stream = new PassThrough();
stream.end(Buffer.from(response.audioContent));
return this.encodeStream(stream, options);
}

private async createAudioResource(options: AudioResourceOptions): Promise<AudioResource<Track>> {
// We want to fall through if play-dl doesn't work, or the speed option is provided
/* eslint-disable no-fallthrough */
switch (this.variant) {
case TrackVariant.TEXT: {
try {
const resource = await googleTextToSpeech();
const resource = await this.googleTextToSpeech(options);
return resource;
} catch (err) {
error(err);
return openAiTextToSpeech();
return this.openAiTextToSpeech(options);
}
}
case TrackVariant.YOUTUBE_LIVESTREAM:
case TrackVariant.YOUTUBE_VOD: {
// play-dl is currently broken
// if (!speed) {
// try {
// const audioResource = await tryPlayDl();
// const audioResource = await this.tryPlayDl(options);
// return audioResource;
// } catch (err) {
// error('Error playing resource from play-dl', err);
// }
// }
}
default: {
return tryYtdlExec();
return this.tryYtdlExec(options);
}
}
/* eslint-enable no-fallthrough */
Expand Down

0 comments on commit d17ac76

Please sign in to comment.