Skip to content
This repository was archived by the owner on Jun 11, 2026. It is now read-only.

OpenVoiceOS/ovos-stt-plugin-HiTZ

Repository files navigation

This repository is archived. No further updates will be made. The same Basque models are available via lighter ONNX-based plugins. See the migration guide below.

OVOS HiTZ STT

OpenVoiceOS STT plugin for Basque models trained by HiTZ.

GPU is strongly recommended for the NeMo path; the ONNX path runs well on CPU.

Online demo: HiTZ/Aholab's Basque Speech-to-Text

Migration Guide

Option A — ONNX (recommended, no PyTorch)

Uses the ONNX exports of the same HiTZ models via ovos-stt-plugin-onnx-asr. Models are part of the OpenVoiceOS STT ONNX collection.

pip install ovos-stt-plugin-onnx-asr
"stt": {
    "module": "ovos-stt-plugin-onnx-asr",
    "ovos-stt-plugin-onnx-asr": {
        "model": "OpenVoiceOS/stt-eu-conformer-ctc-large-onnx"
    }
}
Model HuggingFace Architecture
OpenVoiceOS/stt-eu-conformer-ctc-large-onnx link Conformer-CTC
OpenVoiceOS/stt-eu-conformer-transducer-large-onnx link Conformer-Transducer

Option B — NeMo (original path, requires PyTorch + CUDA)

pip install ovos-stt-plugin-nemo
"stt": {
    "module": "ovos-stt-plugin-nemo",
    "ovos-stt-plugin-nemo": {
        "model": "stt_eu_conformer_transducer_large",
        "lang": "eu"
    }
}
Model HuggingFace
stt_eu_conformer_transducer_large HiTZ/stt_eu_conformer_transducer_large
stt_eu_conformer_ctc_large HiTZ/stt_eu_conformer_ctc_large

Credits

This plugin was developed by TigreGotico for OpenVoiceOS under the ILENIA project.

This plugin was funded by the Ministerio para la Transformación Digital y de la Función Pública and Plan de Recuperación, Transformación y Resiliencia - Funded by EU – NextGenerationEU within the framework of the project ILENIA with reference 2022/TL22/00215337

HiTZ/Aholab's Basque Speech-to-Text model Conformer — models trained on a composite dataset comprising 548 hours of Basque speech, fine-tuned from a pre-trained Spanish stt_es_conformer_ctc_large model. Non-autoregressive "large" Conformer variant, ~121 million parameters.

This project with reference 2022/TL22/00215335 has been partially funded by the Ministerio de Transformación Digital and by the Plan de Recuperación, Transformación y Resiliencia – Funded by the European Union – NextGenerationEU ILENIA and by the project IkerGaitu funded by the Basque Government. This model was trained at Hyperion, one of the high-performance computing (HPC) systems hosted by the DIPC Supercomputing Center.

Releases

Sponsor this project

Packages

Used by

Contributors

Languages