API reference

Status#

Quickstart#

One POST returns a complete WAV. Set stream to true and you get audio frames as each sentence finishes instead.

curl https://api.strophic.in/v1/speech \
  -H "Authorization: Bearer $VOICER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "The deploy finished in ninety seconds.",
    "voice": "af_heart",
    "speed": 1.0,
    "format": "wav"
  }' --output speech.wav

Authentication#

Keys are bearer tokens scoped to one account. They carry the same credit balance as the Studio — there is no separate API quota.

Authorization: Bearer vk_live_a1b2c3…

Keys are shown once at creation and stored only as a hash. Send them from a server — a key in browser code is a key someone else is using.

Create speech#

A successful call returns audio/wav at 24 kHz mono, 16-bit. With format: "pcm" you get raw little-endian int16 with no header.

Streaming format#

With stream: true the response is a sequence of length-prefixed binary frames. Each frame is a one-byte type, a four-byte little-endian payload length, then the payload.

const response = await fetch("https://api.strophic.in/v1/speech", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${key}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ text, voice: "af_heart", stream: true }),
});

const reader = response.body!.getReader();
let buffer = new Uint8Array(0);

while (true) {
  const { done, value } = await reader.read();
  if (done) break;

  buffer = concat(buffer, value);

  // 1 byte frame type, 4 bytes little-endian length, then the payload.
  while (buffer.length >= 5) {
    const view = new DataView(buffer.buffer, buffer.byteOffset);
    const type = view.getUint8(0);
    const length = view.getUint32(1, true);
    if (buffer.length < 5 + length) break;

    const payload = buffer.subarray(5, 5 + length);
    if (type === 2) playPcm16(payload);   // audio
    buffer = buffer.slice(5 + length);
  }
}

List voices#

Returns all 51 voices across 9 languages with their gender and quality grade. The catalogue is static — cache it.

{
  "sample_rate": 24000,
  "languages": [
    { "code": "en-us", "label": "English (US)", "prefix": "a" }
  ],
  "voices": [
    { "id": "af_heart", "name": "Heart", "gender": "female",
      "grade": "A", "lang": "en-us" }
  ]
}

Credits#

One credit is one character of input text. Credits are held when the request is accepted and settled against the audio that actually rendered — if a stream is cut short, the unused balance returns automatically.

The response carries X-Voicer-Credits-Charged and X-Voicer-Credits-Remaining so you can track spend without a second call.

Errors#

Errors are JSON with an error code and a human-readable message. Codes are stable; messages are not.

Rate limits#

Beta accounts are capped at two concurrent generations and thirty requests a minute. The render queue holds eight jobs; past that you get a 503 busy and should retry with backoff.

These numbers exist because Voicer runs on shared hardware, not because we want them to. They will move up as capacity does.