Neural HMMs are all you need (for high-quality attention-free TTS)

[TanUkkii007]

link: https://arxiv.org/abs/2108.13320