This repository contains a complete pipeline for training, preprocessing, and exporting a custom Text-to-Speech (TTS) model. It covers data preparation, model training, and ONNX export for deployment purposes.
| File | Description |
|---|---|
TTS_dataset_preprocess.ipynb |
Preprocesses the audio and text data into a format suitable for training |
TTS_custom_training.ipynb |
Trains the TTS model using a specified configuration and dataset |
TTS_export.ipynb |
Exports the trained model to ONNX format for inference or deployment |
- Preprocessing of raw datasets into JSONL format
- Configuration-driven model training using PyTorch
- ONNX export for runtime-efficient deployment
- Modular and customizable pipeline
Make sure to install the Python dependencies especially all dependencies in the depend.py.
You can find out the models at https://huggingface.co/rhasspy/piper-voices or https://github.com/rhasspy/piper/blob/master/VOICES.md.