GUIDE

A Supertonic 3 API: hosted in the cloud or on your own server

Supertonic 3 is an open text to speech model: small, fast on an ordinary CPU and fluent in 31 languages. It is published as model files, so to call it from an app over HTTP you need a server that runs it. There are two ways: use a hosted API, or run the server yourself.

SuperClonica offers both. Its speech API runs Supertonic 3 voices on our workers, and its free engine runs the same model on your machine, about 1.6 times faster on a processor and 1.7 times faster on a graphics card than the stock code in our measurements.

Get a free API key Hear Supertonic 3 voices Or run the engine yourself

Option 1: the hosted API

Sign in to SuperClonica with Google or GitHub, create a key in the Speech API section of your workspace and send the text:

curl -o hello.ogg https://superclonica.com/api/v1/speech -H "Authorization: Bearer $SUPERCLONICA_KEY" -H "Content-Type: application/json" -d '{"text": "Hello from Supertonic 3!", "voice": "F1", "language": "en"}'

The answer is the audio itself, ogg by default, mp3 or wav on request; with "stream": true it plays from 300 ms in our measurements, whatever the length of the text.

  • Free: a minute of audio a day.
  • Starter, $5 a month: up to 216 hours a month used without a break, and every catalog voice.
  • Voices: the ten stock voices, catalog voices made by authors and your own trained voices.
  • Up to 20,000 characters per request; long texts are voiced in pieces on several workers at once.

Option 2: your own server

The SuperClonica engine runs as a free Docker server with an HTTP API of its own, with no key and no quota:

docker run -d -p 8000:8000 -v superclonica-models:/models superclonica/superclonica:cpu

The first start downloads the model; then POST /synthesize answers with WAV. The guide on running a text to speech API in Docker covers settings, keys and production habits.

Which one to choose

Choose the hosted API when you want to start in minutes, have no server or GPU, need streaming for long texts, or want catalog voices without managing files. Choose your own server when texts must not leave your infrastructure, when you voice so much that your own hardware is cheaper, or when you need to work offline. The voices and the request shape are close, so moving between them later is a small change.

Supertonic 3 limits to know

Supertonic 3 speaks 31 languages but not Chinese yet. Its license, OpenRAIL-M, allows commercial use within its restrictions, including rules on deceptive use and on voices of real people. New voices are style files rather than zero-shot clones: on SuperClonica you can train one from your own recording or pick one in the catalog.