EATS

A pytorch implementation of the EATS: End-to-End Adversarial Text-to-Speech (https://arxiv.org/pdf/2006.03575.pdf)

Attention

• I only implemented the preliminary framework.
• In my opinion, the training cost of EETS is very expensive. Maybe we can consider using ground truth duration during training process and introduce additional auxiliary loss to accelerate convergence.

Author

Geng Yang (@yanggeng1995) Jian Cong (@npujcong)