| FastPitch voices (NVIDIA FastSpeech2-style, text->80-ch mel) for [phoonnx](https://github.com/TigreGotico/phoonnx). Two-stage; voiced via a vocoder. One voice per subfolder (`model.onnx` + native `config.json`). Reference: multi-speaker Arabic FastPitch from [nipponjo/tts_arabic](https://github.com/nipponjo/tts_arabic) (mantoq/buckwalter, paired with tts-arabic-hifigan). Part of the [phoonnx community mirror](https://huggingface.co/collections/OpenVoiceOS/phoonnx-community-mirror-6a2204d302e0e7497870cd68). |
