Neural HMMs are all you need (for high-quality attention-free TTS). (arXiv:2108.13320v5 [eess.AS] UPDATED)

Neural sequence-to-sequence TTS has achieved significantly better output
quality than statistical speech synthesis using HMMs. However, neural TTS is
generally not probabilistic and the use of non-monotonic attention both
increases training time and introduces "babbling" failure modes that are
unacceptable in production. This paper demonstrates that the old and new
paradigms can be combined to obtain the advantages of both worlds. In
particular, we replace the attention in Tacotron 2 with an autoregressive
left-right no-skip hidden Markov model defined by …

arxiv attention