7. Full JavaScript Example

Complete boilerplate to connect, capture microphone input, stream audio, configure a persona, and play back synthesized speech:

// ─────────────────────────────────────────────────────────────
// 1. Queue-based player for streaming 16-bit PCM (24kHz)
// ─────────────────────────────────────────────────────────────
class QueueAudioPlayer {
  constructor() {
    this.audioCtx = new (window.AudioContext || window.webkitAudioContext)();
    this.sampleRate = 24000;
    this.nextStartTime = 0;
  }

  playChunk(int16Array) {
    const float32 = new Float32Array(int16Array.length);
    for (let i = 0; i < int16Array.length; i++) {
      float32[i] = int16Array[i] / 32768.0;
    }
    const buffer = this.audioCtx.createBuffer(1, float32.length, this.sampleRate);
    buffer.getChannelData(0).set(float32);

    const source = this.audioCtx.createBufferSource();
    source.buffer = buffer;
    source.connect(this.audioCtx.destination);

    const now = this.audioCtx.currentTime;
    const playTime = Math.max(now, this.nextStartTime);
    source.start(playTime);
    this.nextStartTime = playTime + buffer.duration;
  }
}

// ─────────────────────────────────────────────────────────────
// 2. Connect and configure the S2S session
// ─────────────────────────────────────────────────────────────
async function startSpeechToSpeech() {
  const player = new QueueAudioPlayer();

  const ws = new WebSocket("ws://localhost:8000/s2s");
  ws.binaryType = "arraybuffer";

  ws.onopen = () => {
    // Configure voice + persona on connection
    // All fields are optional — omit any you don't need
    ws.send(JSON.stringify({
      type:    "config",
      voice:   "alba",             // see GET /voices for full list
      persona: "customer_service"  // or "assistant", "agent", "spiritual"
    }));

    // --- OR use a fully custom system prompt ---
    // ws.send(JSON.stringify({
    //   type:          "config",
    //   voice:         "jean",
    //   persona:       "custom",
    //   system_prompt: "You are a no-nonsense financial advisor. Answer in exactly one sentence."
    // }));
  };

  ws.onmessage = (event) => {
    if (event.data instanceof ArrayBuffer) {
      // Binary frame: synthesized audio chunk (Int16 PCM 24kHz)
      player.playChunk(new Int16Array(event.data));
    } else {
      // Text frame: status, transcription, llm_text, done, or error
      const msg = JSON.parse(event.data);

      if (msg.type === "transcription") console.log("You said:", msg.text);
      if (msg.type === "llm_text")      console.log("Assistant:", msg.text);
      if (msg.type === "error")         console.warn("S2S Error:", msg.message);
      if (msg.type === "done")          console.log("--- Turn complete ---");
    }
  };

  // ─────────────────────────────────────────────────────────
  // 3. Capture microphone and stream Float32 PCM at 16kHz
  // ─────────────────────────────────────────────────────────
  const stream = await navigator.mediaDevices.getUserMedia({
    audio: { channelCount: 1, sampleRate: 16000, echoCancellation: true, noiseSuppression: true }
  });

  const micCtx   = new AudioContext({ sampleRate: 16000 });
  const source    = micCtx.createMediaStreamSource(stream);
  const processor = micCtx.createScriptProcessor(2048, 1, 1);

  processor.onaudioprocess = (e) => {
    if (ws.readyState === WebSocket.OPEN) {
      ws.send(e.inputBuffer.getChannelData(0).buffer); // Float32Array
    }
  };

  source.connect(processor);
  processor.connect(micCtx.destination);
}

startSpeechToSpeech();

8. Standalone Text-to-Speech (REST API)

If you only need to synthesize text into speech without the full speech-to-speech loop, Pocket-S2S exposes standalone GET and POST REST endpoints. Both endpoints return a standard audio/wav byte stream (24kHz, 16-bit, mono PCM) that can be played or saved directly.

<!-- Example of direct embedding in HTML -->
<audio src="http://localhost:8000/tts?text=Hello+from+pocket+TTS&voice=alba" controls></audio>

// Request JSON Body Schema
{
  "text": "The quick brown fox jumps over the lazy dog.",
  "voice": "jean" // optional, defaults to "cosette"
}

async function playStandaloneTTS(text, voice = "cosette") {
  const response = await fetch("/tts", {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({ text, voice })
  });

  if (!response.ok) {
    const err = await response.json();
    console.error("TTS Failed:", err);
    return;
  }

  // Get audio binary data (WAV)
  const blob = await response.blob();
  const audioUrl = URL.createObjectURL(blob);
  
  // Play immediately
  const audio = new Audio(audioUrl);
  audio.play();
}

// Usage:
// playStandaloneTTS("Welcome to the standalone voice synthesizer.", "vera");