phoonnx_ast_miro_unicode

A text-to-speech voice model for Asturian (ast) using the phoonnx VITS architecture, exported to ONNX for efficient CPU inference. The voice persona is Miro — a male speaker adapted via voice conversion from a Galician base model (phoonnx_gl-ES_miro_unicode).

Training data

Trained on TigreGotico/tts_vc_mcv-scripted-v23.0_ast_miro, a synthetic voice-conversion dataset derived from Mozilla Common Voice scripted prompts (release 23.0). Audio was synthesised with a donor TTS voice and then converted to the Miro speaker identity using voice-conversion techniques, producing a multi-hour single-speaker corpus for Asturian.

Usage

import wave
from phoonnx.config import SynthesisConfig
from phoonnx.voice import TTSVoice

voice = TTSVoice.from_hf("OpenVoiceOS/phoonnx_ast_miro_unicode")

synthesis_config = SynthesisConfig(noise_scale=0.667, length_scale=1.0, noise_w_scale=0.8)

with wave.open("output.wav", "wb") as wav_file:
    voice.synthesize_wav("Bones díes, ¿cómo tas?", wav_file, synthesis_config)

Or via the OVOS TTS plugin:

{
  "tts": {
    "module": "ovos-tts-plugin-phoonnx",
    "ovos-tts-plugin-phoonnx": {
      "voice": "OpenVoiceOS/phoonnx_ast_miro_unicode"
    }
  }
}