Streaming TTS, built for developers

// Step 1: authorize your key. This checks your billing/free-tier status
// and returns a short-lived token plus the actual worker URL to connect to.
const { token, url } = await fetch("https://api.readaloudai.org/tts/authorize", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ key: "YOUR_API_KEY" }),
}).then((r) => r.json())

// Step 2: connect DIRECTLY to the worker with that token — this is the
// fast path, no extra relay hop, comparable latency to going straight to
// the GPU worker itself.
const ws = new WebSocket(`${url}?token=${token}`)

ws.onopen = () => {
  ws.send(JSON.stringify({
    type: "synthesize",
    text: "Hello, this is realtime text to speech.",
    voice: "af_heart",
    speed: 1.0,
  }))
}

// Server alternates: a JSON "chunk_meta" frame, immediately followed by
// one binary PCM16LE mono 24kHz audio frame — repeated until "done".
ws.onmessage = (event) => {
  if (typeof event.data === "string") {
    const msg = JSON.parse(event.data)
    if (msg.type === "done") console.log("synthesis complete")
    if (msg.type === "error") console.error(msg.message)
    // msg.type === "chunk_meta" -> next binary frame is this chunk's audio
  } else {
    // binary PCM16LE mono 24kHz frame — queue it for playback
    playAudioChunk(event.data)
  }
}

// The token expires 60 seconds after issuance — call /tts/authorize again
// for each new session rather than trying to reuse one.