This is a notebook from kaggle I had made that allows user's to refine and clean up their tacotron 2 models in terms of the audio output. In order to use this notebook you'll require a Tacotron 2 model fully trained, with the dataset used and the transcription and/or validation file too.