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

arXiv: https://arxiv.org/abs/2108.13320 · HTML: https://mlsanity.com/paper/2108.13320/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 51 references that are in the map.
- First posted on arXiv 30 Aug 2021.

## Builds on (most cited first, 22 of 22)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [WaveNet: A Generative Model for Raw Audio](https://mlsanity.com/paper/1609.03499/) (2016)
- [Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictions](https://mlsanity.com/paper/1712.05884/) (2017)
- [Tacotron: Towards End-to-End Speech Synthesis](https://mlsanity.com/paper/1703.10135/) (2017)
- [Attention-Based Models for Speech Recognition](https://mlsanity.com/paper/1506.07503/) (2015)
- [WaveGlow: A Flow-based Generative Network for Speech Synthesis](https://mlsanity.com/paper/1811.00002/) (2018)
- [Glow-TTS: A Generative Flow for Text-to-Speech via Monotonic Alignment Search](https://mlsanity.com/paper/2005.11129/) (2020)
- [Fixup Initialization: Residual Learning Without Normalization](https://mlsanity.com/paper/1901.09321/) (2019)
- [A Survey on Neural Speech Synthesis](https://mlsanity.com/paper/2106.15561/) (2021)
- [End-to-End Adversarial Text-to-Speech](https://mlsanity.com/paper/2006.03575/) (2020)
- [Non-Attentive Tacotron: Robust and Controllable Neural TTS Synthesis Including Unsupervised Duration Modeling](https://mlsanity.com/paper/2010.04301/) (2020)
- [Efficiently Trainable Text-to-Speech System Based on Deep Convolutional Networks with Guided Attention](https://mlsanity.com/paper/1710.08969/) (2017)
- [Robust Sequence-to-Sequence Acoustic Modeling with Stepwise Monotonic Attention for Neural TTS](https://mlsanity.com/paper/1906.00672/) (2019)
- [Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling](https://mlsanity.com/paper/2103.14574/) (2021)
- [AlignTTS: Efficient Feed-Forward Text-to-Speech System without Explicit Alignment](https://mlsanity.com/paper/2003.01950/) (2020)
- [Unsupervised Neural Hidden Markov Models](https://mlsanity.com/paper/1609.09007/) (2016)
- [Online Segment to Segment Neural Transduction](https://mlsanity.com/paper/1609.08194/) (2016)
- [Speaking Speed Control of End-to-End Speech Synthesis using Sentence-Level Conditioning](https://mlsanity.com/paper/2007.15281/) (2020)
- [End-to-End Text-to-Speech using Latent Duration based on VQ-VAE](https://mlsanity.com/paper/2010.09602/) (2020)
- [Normalizing Flow based Hidden Markov Models for Classification of Speech Phones with Explainability](https://mlsanity.com/paper/2107.00730/) (2021)
- [Median-Based Generation of Synthetic Speech Durations using a Non-Parametric Approach](https://mlsanity.com/paper/1608.06134/) (2016)

## Built on by (most cited first, 8 of 8)

- [Matcha-TTS: A fast TTS architecture with conditional flow matching](https://mlsanity.com/paper/2309.03199/) (2023)
- [Diff-TTSG: Denoising probabilistic integrated speech and gesture synthesis](https://mlsanity.com/paper/2306.09417/) (2023)
- [OverFlow: Putting flows on top of neural transducers for better TTS](https://mlsanity.com/paper/2211.06892/) (2022)
- [Prosody-controllable spontaneous TTS with neural HMMs](https://mlsanity.com/paper/2211.13533/) (2022)
- [Learning-to-Defer in Non-Stationary Time Series via Switching State-Space Models](https://mlsanity.com/paper/2601.22538/) (2026)
- [Fake it to make it: Using synthetic data to remedy the data shortage in joint multimodal speech-and-gesture synthesis](https://mlsanity.com/paper/2404.19622/) (2024)
- [R-MelNet: Reduced Mel-Spectral Modeling for Neural TTS](https://mlsanity.com/paper/2206.15276/) (2022)
- [Pyramidal Hidden Markov Model For Multivariate Time Series Forecasting](https://mlsanity.com/paper/2310.14341/) (2023)

## Abstract (arXiv)

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 uses non-monotonic attention. Attention failures increase training time and can make synthesis babble incoherently. This paper describes how the old and new paradigms can be combined to obtain the advantages of both worlds, by replacing attention in neural TTS with an autoregressive left-right no-skip hidden Markov model defined by a neural network. Based on this proposal, we modify Tacotron 2 to obtain an HMM-based neural TTS model with monotonic alignment, trained to maximise the full sequence likelihood without approximation. We also describe how to combine ideas from classical and contemporary TTS for best results. The resulting example system is smaller and simpler than Tacotron 2, and learns to speak with fewer iterations and less data, whilst achieving comparable naturalness prior to the post-net. Our approach also allows easy control over speaking rate.