Aholab TTS Synthesis models

Description

This repository contains the necessary tools to synthesize speech for Basque (eu), Galician (gl), Catalan (ca) and Spanish (es). The voices used are all based on the VITS architecture proposed by Kim et al. (2021).
Disclaimer: The TTS voices are not uploaded to this Github repository, they are uploaded to a HuggingFace repo. The download of the voices is done automatically when the synthesize.py script is executed. Some of the voices below might not have been uploaded yet.

• Basque [eu]:
  • antton
  • maider
• Galician [gl]:
  • brais
  • celtia
  • iago
  • icia
  • paulo
  • sabela
• Catalan [ca]:
  • pau
  • ona
• Spanish [es]:
  • laura
  • alejandro

Uses

These models are intented to be used for speech synthesis in Basque, Galician, Catalan and Spanish.

Installation

You will need to install huggingface_hub package with the following command.

pip install huggingface_hub

How to use

Python

Use the synthesize.py script to generate speech. All available models are listed in the sections above. Before running the script, navigate to the repository directory:

cd aHoTTS

For help:

python3 synthesize.py -h

Example commands:

python3 synthesize.py -t "Antton naiz, zer moduz zaude." -l eu -m antton -o audio_name  
python3 synthesize.py -t "Soy Laura, qué tal estás?" -l es -m laura -o audio_name  
python3 synthesize.py -t "Sóc Ona, com estàs." -l ca -m ona -o audio_name
python3 synthesize.py -t "Son Brais, como estás." -l gl -m brais -o audio_name

The synthesized audio is saved as a .wav file inside the output/ directory.
Use --output / -o to specify the filename.

Additional information

Voice Resource Licenses and references

• Galician
  • Celtia
    Public Creative Commond Attribution 4.0 International License
    Vázquez Abuín, M., García Díaz, N., Vladu, A. I., Magariños, C., Vidal Miguéns, A., & Fernández Rei, E. (2023). Nos_Celtia-GL: Galician TTS corpus (1.0.0.) [Data set]. Zenodo.
  • Brais
    Public Creative Commond Attribution 4.0 International License
    Vladu, A. I., García Díaz, N., Regueira Fernández, X. L., Magariños, C., Moscoso Sánchez, A., Fernández López, D., Fernández Rei, E., & Dubert-García, F. (2025). Nos_Brais-GL: Galician TTS corpus [Data set]. Zenodo
  • Sabela/Icia/Iago/Paulo
    Public Creative Commond Attribution 4.0 International License
    Centro Ramón Piñeiro para a Investigación en Humanidades (CRPIH), & Multimedia Technology Group (GTM) – atlanTTic Research Center for Telecommunication Technologies. (2023). CRPIH_UVigo-GL-Voices: Galician TTS dataset (1.0.0.) [Data set]. Zenodo.
• Catalan
  • Creative Commons Attribution-ShareAlike 4.0 International Public License festcat_trimmed_denoised
• Basque
  • Maider, Antton: developed by HiTZ with funding from Project ILENIA. Public Creative Commond Attribution 4.0
• Spanish
  • Alejandro: Developed in HiTZ from openSLR dataset.
  • Laura: Acquired in ELRA ID: ELRA-S0309

Authors

HiTZ Basque Center for Language Technology - Aholab Signal Processing Laboratory, University of the Basque Country EHU.

Contact information

Ibon Saratxaga: ibon.saratxaga@ehu.eus

Licensing Information

Apache License, Version 2.0

Funding

Catalan and Galician have been funded by the project with reference numbers 2022/TL22/00215337, 2022/TL22/00215336, 2022/TL22/00215335, and 2022/TL22/00215334 is funded by the Ministry of Digital Transformation and by the Recovery, Transformation and Resilience Plan – Funded by the European Union – NextGenerationEU.

Citation information

García, V., Hernáez, I., & Navas, E. (2022). Evaluation of Tacotron Based Synthesizers for Spanish and Basque. Applied Sciences, 12(3), 1686. https://doi.org/10.3390/app12031686