Microsoft's AI generates realistic speech with only 200 training samples

In a newly published paper, Microsoft researchers describe a state-of-the-art AI speech system that needs only 200 samples to achieve high accuracy.