For training a TTS model, you need a dataset with speech recordings and transcriptions. See the following pages for more information on:
:maxdepth: 1 formatting_your_dataset what_makes_a_good_dataset tts_datasets