Get Started#

This page guides you through the fastest path to a running Text To Speech microservice. The recommended deployment uses Docker Compose. Alternative deployment options are available in the How-to Guides section.

Before You Begin#

Deploy with Docker#

The container image exposes the API on host port 8011 and mounts shared folders for models, storage, and the Hugging Face cache.

docker compose up -d --build

If you hit permission errors on models/, storage/, or .cache/huggingface/, see Troubleshooting.

For the full step-by-step guide, see Run with Docker Compose.

Verify#

Once the service is running:

curl --noproxy '*' http://127.0.0.1:8011/health

Expected response:

{"status": "ok"}

Try It Out#

Once the service responds to the health check, send a speech synthesis request:

curl --noproxy '*' -sS \
  -o speech.wav \
  -w '%{http_code}\n' \
  -X POST http://127.0.0.1:8011/v1/audio/speech \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "default",
    "input": "The kiosk is ready for your next request.",
    "response_format": "wav"
  }'

Expected output:

200

The synthesized audio is saved to speech.wav in the current directory. Play it with any WAV-capable player (aplay speech.wav on Linux, or open it in a media player).

To list available voices and confirm the active model:

curl --noproxy '*' http://127.0.0.1:8011/v1/audio/voices

Expected output (example with the default SpeechT5 model):

{
  "model": "microsoft/speecht5_tts",
  "runtime": "openvino",
  "speakers": ["default"],
  "languages": ["English"]
}

Note: First startup may take longer than usual because the model is downloaded and converted during initialization. Subsequent starts are faster.

Next Steps#

Other Deployment Options#