Skip to navigation

Quickstart

Make your first speech requests in a few minutes: one call that turns text into audio, one that turns that audio back into text, and one that shows what they cost. The examples use Python with the requests library, or Node.js 18 or newer.

1

Create an API key

Sign up at api.llynx.io with your business email, verify it and accept the terms. Open API keys and create a key. It is shown once; store it on your server and never in a browser or mobile app. The examples read it from the LLYNX_API_KEY environment variable.

2

Turn text into speech

Send the text, its language tag and an optional voice gender. The response body is a complete WAV file, so save it straight to disk.

import os
import requests
response = requests.post(
    "https://api.llynx.io/v2/audio/speech",
    headers={"Authorization": f"Bearer {os.environ['LLYNX_API_KEY']}"},
    json={
        "input": "Your order has shipped and will arrive on Thursday.",
        "language": "en",
        "gender": "female",
        "response_format": "wav",
    },
)
response.raise_for_status()
with open("hello.wav", "wb") as f:
    f.write(response.content)
print(response.headers["X-Llynx-Credits-Reserved"], "credits")

Play hello.wav. This sentence is 51 characters, so it costs about 0.5 credits; the X-Llynx-Credits-Reserved response header shows the exact figure.

3

Turn speech into text

Upload a recording with the language spoken in it. Here bonjour.m4a is a short French voice note; the hello.wav you just generated works too, with language set to en.

import os
import requests
with open("bonjour.m4a", "rb") as audio:
    response = requests.post(
        "https://api.llynx.io/v2/audio/transcriptions",
        headers={"Authorization": f"Bearer {os.environ['LLYNX_API_KEY']}"},
        files={"file": ("bonjour.m4a", audio, "audio/mp4")},
        data={"language": "fr"},
    )
response.raise_for_status()
result = response.json()
print(result["text"])

The response is JSON:

{
  "id": "req_7c1d9b2e-4f3a-4b8e-9c0d-2a6e5f1b3d47",
  "text": "Bonjour, votre commande arrivera jeudi.",
  "language": "fr",
  "language_mode": "explicit",
  "duration_ms": 2800,
  "capabilities_revision": "v1",
  "usage": {
    "credit_unit": "stt_credit",
    "credits_reserved": 0.15,
    "billable_seconds": 3
  }
}

Any recording works the same way: WAV, MP3, M4A, FLAC, Ogg or WebM, up to 60 seconds. Swap the tag to try another language, for example hi, ta or es.

4

Check your balance

import os
import requests
response = requests.get(
    "https://api.llynx.io/v2/audio/credits",
    headers={"Authorization": f"Bearer {os.environ['LLYNX_API_KEY']}"},
)
response.raise_for_status()
balances = response.json()
print(balances["stt"]["available"], "STT credits,", balances["tts"]["available"], "TTS credits")

New accounts start with introductory credit for each service. Balances for speech to text and text to speech are separate; the response shows allocated, used and available credit for each.

Before you go to production

  • Languages. Every request names its language with a tag from the Languages page. Requests in other languages are rejected.
  • Limits. Recordings up to 60 seconds and 32 MiB; text up to 2,000 characters per request. Generated audio is WAV, 16-bit mono at 24 kHz.
  • Charges. Transcription costs one credit per 20 seconds of audio, rounded up to the whole second. Speech costs one credit per 100 input characters. Buy more with a one-time pack; see credits and packs.
  • Failures. A request that fails after it was accepted is refunded. Results are not stored, so a retry is a new request; see retries and failures.
  • Reference. Request and response fields, headers and error codes are in the API reference.