Text-To-Speech (TTS) using Gemini API with Google Apps Script

/**
 * Converts a byte data of "audio/L16" to a byte data of "audio/wav".
 * L16 assumes 16-bit PCM.
 *
 * @param {Byte[]} inputData Input data (audio/L16).
 * @param {string} mimeType
 * @param {number} numChannels - Mono and stereo are 1 and 2, respectively.
 * @return {Byte[]} Converted data as byte data.
 */
function convertL16ToWav_(inputData, mimeType = "audio/L16;codec=pcm;rate=24000", numChannels = 1) {
  const [type, codec, sampleRate] = mimeType.split(";").map(e => e.includes("=") ? e.trim().split("=")[1] : e.trim());
  if (type != "audio/L16" || codec != "pcm") {
    throw new Error(`Sorry. This function can use for only "audio/L16;codec=pcm;rate=###".`);
  }
  const bitsPerSample = 16;
  const blockAlign = numChannels * bitsPerSample / 8;
  const byteRate = Number(sampleRate) * blockAlign;
  const dataSize = inputData.length;
  const fileSize = 36 + dataSize;
  const header = new ArrayBuffer(44);
  const view = new DataView(header);
  const data = [
    { method: "setUint8", value: [..."RIFF"].map(e => e.charCodeAt(0)), add: [0, 1, 2, 3] },
    { method: "setUint32", value: [fileSize], add: [4], littleEndian: true },
    { method: "setUint8", value: [..."WAVE"].map(e => e.charCodeAt(0)), add: [8, 9, 10, 11] },
    { method: "setUint8", value: [..."fmt "].map(e => e.charCodeAt(0)), add: [12, 13, 14, 15] },
    { method: "setUint32", value: [16], add: [16], littleEndian: true },
    { method: "setUint16", value: [1, numChannels], add: [20, 22], littleEndian: true },
    { method: "setUint32", value: [Number(sampleRate), byteRate], add: [24, 28], littleEndian: true },
    { method: "setUint16", value: [blockAlign, bitsPerSample], add: [32, 34], littleEndian: true },
    { method: "setUint8", value: [..."data"].map(e => e.charCodeAt(0)), add: [36, 37, 38, 39] },
    { method: "setUint32", value: [dataSize], add: [40], littleEndian: true },
  ];
  data.forEach(({ method, value, add, littleEndian }) =>
    add.forEach((a, i) => view[method](a, value[i], littleEndian || false))
  );
  return [...new Uint8Array(header), ...inputData];
}

function myFunction() {
  const apiKey = "###"; // Please set your API key here.

  const text = [
    "Create Text-To-Speech the following conversation.",
    "User A: Hey there! How are you doing today?",
    "User B: Hi! I'm doing well, thanks. How about you?",
    "User A: I'm good too, thanks for asking!",
  ].join("\n");
  const url = `https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-preview-tts:generateContent?key=${apiKey}`;
  const payload = {
    contents: [{ role: "user", parts: [{ text }] }],
    generationConfig: {
      responseModalities: ["AUDIO"],
      speechConfig: {
        multiSpeakerVoiceConfig: {
          speakerVoiceConfigs: [
            {
              speaker: "User A",
              voiceConfig: { prebuiltVoiceConfig: { voiceName: "Kore" } },
            },
            {
              speaker: "User B",
              voiceConfig: { prebuiltVoiceConfig: { voiceName: "Leda" } },
            },
          ],
        },
      },
    },
  };
  const res = UrlFetchApp.fetch(url, {
    contentType: "application/json",
    payload: JSON.stringify(payload),
  });
  const obj = JSON.parse(res.getContentText());
  const { data, mimeType } = obj.candidates[0].content.parts[0].inlineData;
  const blob = Utilities.newBlob(
    convertL16ToWav_(Utilities.base64Decode(data), mimeType),
    "audio/wav",
    "sample.wav"
  );
  DriveApp.createFile(blob);
}