2 months ago, I was halfway into publishing a series on my TTS project. Today, the site for that project is defunct, and no new posts have been made about it.

What happened?

Let’s say, for some odd reason, you’re hosting a Python service that:

1. takes requests that run large computations based on a randomly generated number (a “seed”)
2. spreads work across multiple workers to handle many requests

Then, it’s likely you’ll introduce a subtle randomness bug that leads to duplicate seeds appearing. Let me explain:

With the raw data in tow, we can construct a proper TTS Dataset with the use of a few Python scripts.

Back when 2019’s Advent of Code was going on, I started a habit of what I now call phonecoding – the act of programming using a run-of-the-mill Android smartphone.

To train an AI Text-To-Speech (TTS) model, we’ll need to obtain a Labelled Dataset with two things:

1. Clean audio files, containing only the voice we’re cloning
2. The dialogue transcript (text) for each audio file