Audio samples from "HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis"