Description

This is an OpenVoiceOS TTS plugin for PicoTTS, a fully offline text-to-speech engine.

Install

pip install ovos-tts-plugin-pico

pico2wave must be available. On a Raspberry Pi, install it with:

apt-get install libttspico0
apt-get install libttspico-utils

You can also install it from source.

Configuration

  "tts": {
    "module": "ovos-tts-plugin-pico"
 }

The voice corresponds to a language code. OVOS auto-detects it from the global config.

You can also set it explicitly to one of "de-DE", "es-ES", "fr-FR", "it-IT", "en-US":

  "tts": {
    "module": "ovos-tts-plugin-pico",
    "ovos-tts-plugin-pico": {
      "voice": "en-US"
    }
 }

Docker (ovos-tts-server)

A container image runs the plugin as an ovos-tts-server (ElevenLabs-compatible API). CI builds and pushes the image to GHCR on every push to dev or master:

docker run -p 9666:9666 ghcr.io/openvoiceos/ovos-tts-plugin-pico:latest
curl "http://localhost:9666/synthesize/hello%20world?lang=en-US" --output hello.wav

The PICO_VOICE build arg bakes the served voice into the image (default en-US; valid values de-DE, en-GB, en-US, es-ES, fr-FR, it-IT). Rebuild the image to change it, for example:

docker build --build-arg PICO_VOICE=de-DE -t pico-tts .

See the bundled docker-compose.yml. Pico runs fully offline, so the container needs no network access or API keys.

Related projects

• ovos-plugin-manager loads this plugin through the opm.tts entry point.
• ovos-tts-server serves this plugin over an HTTP API, and is what the bundled Docker image runs.

License

Apache-2.0. See LICENSE.