I'm working in a tool "voicebuilder" that can take an ebook(epub) and an audiobook (e.g. mp3) and build an ljspeech dataset to train a text-to-speech model with

I'm working in a tool "voicebuilder" that can take an ebook(epub) and an audiobook (e.g. mp3) and build an ljspeech dataset to train a text-to-speech model with your favourite narrators voice. It splits the ebook into sentences and matches the audio part for each sentence, similar to whispersync. My goal is to have a minimum of manual steps and it is working out pretty well so far. Open sourcing it is part of the plan, but docs are important and it will take a while...