The Github readme doesn't list this: what data trained this? Was it done by the voices of the creators, or was this trained on data scraped from the internet or

The Github readme doesn't list this: what data trained this? Was it done by the voices of the creators, or was this trained on data scraped from the internet or other archives?