This blog was originally posted in the OpenVoiceOS blog
phoonnx is now the primary Text-to-Speech framework for OpenVoiceOS. It is a training and inference stack built on VITS (a neural text-to-speech model architecture) and ONNX (a portable format for running trained models), and it produces consistent, offline-ready voices across every language we support.
Hear it in action: the Voices Demo runs phoonnx voices live in your browser — onnxruntime-web, no server.
New Language: Introducing Basque!
We built on our previous work on Making Synthetic Voices From Scratch and our Arabic TTS Collaboration to add voices for Basque (eu-ES).
This adds both a male voice (Miro) and a female voice (Dii). It continues our push to support low-resource languages that lack open, high-quality TTS options.
Before this, the only Basque voice was a robotic female voice from the AhoTTS plugin, built in collaboration with ILENIA.
Hear the results: examples of the new Basque open-source voices.
Miro (Male Voice): Listen to Miro
Dii (Female Voice): Listen to Dii
A Consistent Voice Identity Across Languages
OpenVoiceOS needs a consistent brand voice: a standard male and female persona that sounds the same regardless of which language the assistant is configured for. A user who sets up OpenVoiceOS in Lisbon and later switches to German should hear the same familiar speaker.
TigreGotico builds and maintains the phoonnx engine, contributes the open training datasets, and trains the default multilingual OVOS voices that fulfil this commitment: Miro (male) and Dii (female).
Phonemizer flexibility
phoonnx is not only an inference tool. It is a training and inference framework built on the VITS architecture, and it lets us prototype, train, and deploy new voices quickly.
Part of that comes from support for different phonemizers. A phonemizer (or G2P — Grapheme-to-Phoneme model) converts written text into the sequence of sound units (phonemes) the TTS model speaks. Different languages often need different, specialized phonemizers for accurate speech.
- eSpeak compatibility: phoonnx models trained with the widely available eSpeak phonemizer are fully compatible with the Piper TTS engine’s runtime. This means easy deployment inside the existing OVOS ecosystem and in third-party projects like Home Assistant.
- Custom phonemizer support: the framework is not limited to eSpeak. For example, the Galician models built by Proxecto Nós with the Cotovia phonemizer are fully compatible and can be used with the phoonnx pipeline.
This flexibility lets us reuse the work of other open-source projects. For inference, phoonnx can use models originally trained by other projects, including Coqui, Mimic3, and Piper.
Next: ByT5-based G2P models
We keep working to improve G2P accuracy, especially for low-resource languages. We are developing and testing new G2P models based on the ByT5 architecture. These transformer-based models aim to phonemize more accurately across a wider range of languages.
You can follow their development here: G2P Models Collection.
In the near future a dedicated OVOS TTS plugin will be created for phoonnx and made the default for OpenVoiceOS, replacing the previous plugins: ovos-tts-plugin-piper and ovos-tts-plugin-nos.
In the meantime you can try the new voices via the existing plugin ovos-tts-plugin-piper
All you need to do is pass the model urls under mycroft.conf
"tts": {
"module": "ovos-tts-plugin-piper",
"ovos-tts-plugin-piper": {
"model": "https://huggingface.co/OpenVoiceOS/phoonnx_eu-ES_miro_espeak/resolve/main/miro_eu-ES.onnx",
"model_config": "https://huggingface.co/OpenVoiceOS/phoonnx_eu-ES_miro_espeak/resolve/main/miro_eu-ES.piper.json"
}
}Progress Report: Available Languages
The collective work of the OpenVoiceOS and TigreGotico teams has resulted in a rapidly expanding library of open-source TTS models.
Currently Supported Languages:
- Arabic
- Basque
- Dutch
- English (US/GB)
- French
- German
- Italian
- Portuguese (Brazil/Portugal)
- Spanish
Get Involved and Find the Models
We invite the community to explore and use these new resources.
| Resource | Description | Link |
|---|---|---|
| Phoonnx Models | The new phoonnx-trained TTS models in ONNX format. | phoonnx-tts-models |
| Piper/Phoonnx Voices | The full collection of OpenVoiceOS voices compatible with Piper. | pipertts-voices |
| Open Datasets | Datasets used for training these voices, furthering open-data research. | tts-datasets |
Explore the full model and dataset catalogue on Hugging Face: TigreGotico · OpenVoiceOS. To contribute data, open an issue or discussion on phoonnx.