Custom Indonesian VITS Text-to-speech Training with Coqui TTS

• Made by : Zahrizhal Ali

This project uses the VITS (Very Deep Image Prior for Text-to-Speech) model and Coqui TTS (Text-to-Speech) to generate speech from text. The VITS model is a deep learning model that can convert text to mel-spectrogram, which is then used by Coqui TTS to generate speech.

Installation

To install the Text-to-Speech project using VITS model and Coqui TTS, follow these steps:

1. Clone the repository using git clone https://github.com/ZahrizhalAli/indonesian-tts-vits.git
2. Navigate to the project directory using cd indonesian-tts-vits
3. Install the dependencies, if you use google collab its even better
4. Start set up your dataset and transcript

Customization

The Text-to-speech project can be customized by modifying the following parameters:

• sample_rate: The sample rate of the output speech (default is 22050)
• epochs: Training iteration
• save_checkpoints : to save checkpoint in case we need to continue training our interrupted training

Credits

Coqui TTS