Umbra Audio is a system to help create audiobooks from an ePub file.

It is NOT a click a button and get an audio book. There are multiple steps the AI assists with, then I check and validate before going to the next step. The process steps are:

1. Segment the book into chapters, scenes, and utterances. Utterances are some vocalization that can be expressive.
2. The system searches for every speaker and the aliases for that speaker.
   1. Bob, Robert, and Uncle Bob are all the same speaker in different scenes. They all need to have the same voice.
   1. I review the speakers and have the system generate a voice for each.
3. The next step is to review and validate the spoken names. A name could have a very different pronunciation that what a text to speech system would say. So I have to give the system the phonic pronunciation. Examples like Niamh (nee-ev) and Siobhan (shuh-von).
4. Now I can get into the chapters and utterances. The first thing is that the ePub book has chapters or sections that aren’t needed in an audiobook The long text at the beginning can be shortened.
   1. Shorten, remove, alter text at the beginning.
5. Each utterance needs to be checked.
   1. Is the speaker correct?
   1. Is the tone variation good? Neutral, surprised, angry, sad, etc.
   1. The words to speak, are they split properly by the system?
      1. The system can sometimes have a narration and a spoken statement together in the utterance. They must be split.
6. If everything has worked properly to this point, the text to speech system (TTS) will create an audio with the proper voice, saying the proper names correctly, with the proper intonation.
7. Put each utterance together in a chapter, then the chapters get put together and you have an audiobook.

I have been working on the system, application and the book at the same time. While I have published Deal Hunter audiobook I am still working on the Umbra Audio system.