My understanding is that before it’s trained on the actors voice it still requires massive amounts of data for the neural net to understand the rules of human speech. Those people aren’t being compensated or even asked.