This includes the ability to train text-to-speech on a human voice so that you could create recordings that sound as if a specific person spoke them – a technology with obvious potential for abuse, though the company said all customers requesting this will be subject to ethical review. 1

Requests left for this source ( 5