GPT Realtime Translate

{
  "audio": {
    "output": {
      "language": "es"
    }
  }
}

// Node.js 22+. speech.pcm is 24 kHz, 16-bit, mono PCM; make one with:
// ffmpeg -i speech.mp3 -ar 24000 -ac 1 -f s16le speech.pcm
import { readFile, writeFile } from "node:fs/promises";

const socket = new WebSocket("wss://api.ohmygpt.com/v1/realtime/translations?model=openai/gpt-realtime-translate", ["realtime", "openai-insecure-api-key." + process.env.OHMYGPT_API_KEY]);
const voice = [];
socket.onmessage = async (event) => {
  const data = JSON.parse(event.data);
  if (data.type === "session.output_transcript.delta") process.stdout.write(data.delta);
  if (data.type === "session.output_audio.delta") voice.push(Buffer.from(data.delta, "base64"));
  if (data.type === "error") console.error(data.error);
  if (data.type === "session.closed" || data.type === "error") {
    // Play it with: ffplay -f s16le -ar 24000 -ac 1 translation.pcm
    await writeFile("translation.pcm", Buffer.concat(voice));
    socket.close();
  }
};
socket.onopen = async () => {
  socket.send(JSON.stringify({ type: "session.update", session: { audio: { output: { language: "es" } } } }));
  const audio = await readFile("speech.pcm");
  // Send at real-time speed, 100 ms per chunk, as a microphone would
  for (let offset = 0; offset < audio.length; offset += 4800) {
    socket.send(JSON.stringify({ type: "session.input_audio_buffer.append", audio: audio.subarray(offset, offset + 4800).toString("base64") }));
    await new Promise((resolve) => setTimeout(resolve, 100));
  }
  // The server finishes the translation in flight, then sends session.closed
  socket.send(JSON.stringify({ type: "session.close" }));
};