This repository contains code for training a Text-to-Speech (TTS) model specifically for Hindi language using the VITS model. The VITS model is known for its high-quality speech synthesis capabilities.

Installation

git clone https://github.com/Ishank56/vits_using_coqui.tts.git pip install -e .

1. Ensure that the Hindi dataset is available inside the Dataset folder. The Hindi data i used can be downloaded from here. The data should be formatted in a manner similar to LJSpeech_1.1 dataset for compatibility.

Dataset I used for training here: https://keithito.com/LJ-Speech-Dataset/

1. Install all required libraries for phonemizing Hindi alphabets. Espeak library is particularly useful for this purpose. For specific files config.json needs to be set accordingly while using vits model,
2. Adjust the parameters in the code according to your requirements. The rest of the parameters should already be updated accordingly for the Hindi dataset.