Enhancing audio quality for expressive Neural Text-to-Speech

Abstract

Artificial speech synthesis has made a great leap in terms of naturalness as recent Text-to-Speech (TTS) systems are capable of producing speech with similar quality to human recordings. However, not all speaking styles are easy to model: highly expressive voices are still challenging even to recent TTS architectures since there seems to be a trade-off between expressiveness in a generated audio and its signal quality. In this paper, we present a set of techniques that can be leveraged to enhance the signal quality of a highly-expressive voice without the use of additional data. The proposed techniques include: tuning the autoregressive loop's granularity during training; using Generative Adversarial Networks in acoustic modelling; and the use of Variational Auto-Encoders in both the acoustic model and the neural vocoder. We show that, when combined, these techniques greatly closed the gap in perceived naturalness between the baseline system and recordings by 39% in terms of MUSHRA scores for an expressive celebrity voice.

arXiv comment6 pages, 4 figures, 2 tables, SSW 2021

References in corpus (14)

• Auto-Encoding Variational Bayes
• WaveNet: A Generative Model for Raw Audio
• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Autoencoding beyond pixels using a learned similarity metric
• Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
• Improving Variational Inference with Inverse Autoregressive Flow
• Learning Discourse-level Diversity for Neural Dialog Models using Conditional Variational Autoencoders
• Improved Variational Autoencoders for Text Modeling using Dilated Convolutions
• Forward Attention in Sequence-to-sequence Acoustic Modelling for Speech Synthesis
• Unsupervised Learning of Visual Structure using Predictive Generative Networks
• Hierarchical Generative Modeling for Controllable Speech Synthesis
• Avoiding Latent Variable Collapse With Generative Skip Models
• Robust Sequence-to-Sequence Acoustic Modeling with Stepwise Monotonic Attention for Neural TTS
• Parallel WaveNet conditioned on VAE latent vectors