feat(helpers): add new audio helpers

examples/speech-to-text.ts

@@ -0,0 +1,19 @@
+import OpenAI from 'openai';
+import { recordAudio } from 'openai/helpers/audio';
+
+const openai = new OpenAI();
+
+async function main(): Promise<void> {
+  console.log('Recording for 5 seconds...');
+  const response = await recordAudio({ timeout: 5000, device: 4 });
+
+  console.log('Transcribing...');
+  const transcription = await openai.audio.transcriptions.create({
+    file: response,
+    model: 'whisper-1',
+  });
+
+  console.log(transcription.text);
+}
+
+main().catch(console.error);

examples/text-to-speech.ts

@@ -0,0 +1,23 @@
+import OpenAI from 'openai';
+import { playAudio } from 'openai/helpers/audio';
+
+const openai = new OpenAI();
+
+const exampleText = `
+I see skies of blue and clouds of white
+The bright blessed days, the dark sacred nights
+And I think to myself
+What a wonderful world
+`.trim();
+
+async function main(): Promise<void> {
+  const response = await openai.audio.speech.create({
+    model: 'tts-1',
+    voice: 'nova',
+    input: exampleText,
+  });
+
+  await playAudio(response);
+}
+
+main().catch(console.error);

src/helpers/audio.ts

@@ -0,0 +1,145 @@
+import { File } from 'formdata-node';
+import { spawn } from 'node:child_process';
+import { Readable } from 'node:stream';
+import { platform, versions } from 'node:process';
+import { Response } from 'openai/_shims';
+
+const DEFAULT_SAMPLE_RATE = 24000;
+const DEFAULT_CHANNELS = 1;
+
+const isNode = Boolean(versions?.node);
+
+const recordingProviders: Record<NodeJS.Platform, string> = {
+  win32: 'dshow',
+  darwin: 'avfoundation',
+  linux: 'alsa',
+  aix: 'alsa',
+  android: 'alsa',
+  freebsd: 'alsa',
+  haiku: 'alsa',
+  sunos: 'alsa',
+  netbsd: 'alsa',
+  openbsd: 'alsa',
+  cygwin: 'dshow',
+};
+
+function isResponse(stream: NodeJS.ReadableStream | Response | File): stream is Response {
+  return typeof (stream as any).body !== 'undefined';
+}
+
+function isFile(stream: NodeJS.ReadableStream | Response | File): stream is File {
+  return stream instanceof File;
+}
+
+async function nodejsPlayAudio(stream: NodeJS.ReadableStream | Response | File): Promise<void> {
+  return new Promise((resolve, reject) => {
+    try {
+      const ffplay = spawn('ffplay', ['-autoexit', '-nodisp', '-i', 'pipe:0']);
+
+      if (isResponse(stream)) {
+        stream.body.pipe(ffplay.stdin);
+      } else if (isFile(stream)) {
+        Readable.from(stream.stream()).pipe(ffplay.stdin);
+      } else {
+        stream.pipe(ffplay.stdin);
+      }
+
+      ffplay.on('close', (code: number) => {
+        if (code !== 0) {
+          reject(new Error(`ffplay process exited with code ${code}`));
+        }
+        resolve();
+      });
+    } catch (error) {
+      reject(error);
+    }
+  });
+}
+
+export async function playAudio(input: NodeJS.ReadableStream | Response | File): Promise<void> {
+  if (isNode) {
+    return nodejsPlayAudio(input);
+  }
+
+  throw new Error(
+    'Play audio is not supported in the browser yet. Check out https://npm.im/wavtools as an alternative.',
+  );
+}
+
+type RecordAudioOptions = {
+  signal?: AbortSignal;
+  device?: number;
+  timeout?: number;
+};
+
+function nodejsRecordAudio({ signal, device, timeout }: RecordAudioOptions = {}): Promise<File> {
+  return new Promise((resolve, reject) => {
+    const data: any[] = [];
+    const provider = recordingProviders[platform];
+    try {
+      const ffmpeg = spawn(
+        'ffmpeg',
+        [
+          '-f',
+          provider,
+          '-i',
+          `:${device ?? 0}`, // default audio input device; adjust as needed
+          '-ar',
+          DEFAULT_SAMPLE_RATE.toString(),
+          '-ac',
+          DEFAULT_CHANNELS.toString(),
+          '-f',
+          'wav',
+          'pipe:1',
+        ],
+        {
+          stdio: ['ignore', 'pipe', 'pipe'],
+        },
+      );
+
+      ffmpeg.stdout.on('data', (chunk) => {
+        data.push(chunk);
+      });
+
+      ffmpeg.on('error', (error) => {
+        console.error(error);
+        reject(error);
+      });
+
+      ffmpeg.on('close', (code) => {
+        returnData();
+      });
+
+      function returnData() {
+        const audioBuffer = Buffer.concat(data);
+        const audioFile = new File([audioBuffer], 'audio.wav', { type: 'audio/wav' });
+        resolve(audioFile);
+      }
+
+      if (typeof timeout === 'number' && timeout > 0) {
+        const internalSignal = AbortSignal.timeout(timeout);
+        internalSignal.addEventListener('abort', () => {
+          ffmpeg.kill('SIGTERM');
+        });
+      }
+
+      if (signal) {
+        signal.addEventListener('abort', () => {
+          ffmpeg.kill('SIGTERM');
+        });
+      }
+    } catch (error) {
+      reject(error);
+    }
+  });
+}
+
+export async function recordAudio(options: RecordAudioOptions = {}) {
+  if (isNode) {
+    return nodejsRecordAudio(options);
+  }
+
+  throw new Error(
+    'Record audio is not supported in the browser. Check out https://npm.im/wavtools as an alternative.',
+  );
+}