Speech is becoming ever more important in relations between humans and computer-based technologies. Deep learning is deployed to improve voice software. How good is it? In what realms can it be used?