TFGAN: Time and Frequency Domain Based Generative Adversarial Network for High-fidelity Speech Synthesis - View it on GitHub