A ONNX multi-speaker Vits Text-to-Speech model, trained on Ofis Publik Ar Brezhoneg data.

Install sherpa-onnx Python module for inference:

Download files model.onnx and tokens.txt from this repository.

Use --sid=0 for Aziliz voice and --sid=1 for Per/Paol voice.

Stephen Russell and the Language Technologies department, Bangor University