It would be great to be able to generate an AI voice from a real voice sample.