Nana-HDR: A Non-attentive Non-autoregressive Hybrid Model for TTS

TL;DR

Nana-HDR introduces a hybrid non-attentive, non-autoregressive TTS model combining Transformer and RNN components, achieving high naturalness and robustness in Mandarin speech synthesis.

Contribution

It proposes a novel Dense-fuse encoder and a duration predictor to enhance non-autoregressive TTS without attention mechanisms.

Findings

Achieves competitive naturalness in Mandarin TTS

Demonstrates robustness across different datasets

Utilizes hybrid Transformer-RNN architecture effectively

Abstract

This paper presents Nana-HDR, a new non-attentive non-autoregressive model with hybrid Transformer-based Dense-fuse encoder and RNN-based decoder for TTS. It mainly consists of three parts: Firstly, a novel Dense-fuse encoder with dense connections between basic Transformer blocks for coarse feature fusion and a multi-head attention layer for fine feature fusion. Secondly, a single-layer non-autoregressive RNN-based decoder. Thirdly, a duration predictor instead of an attention model that connects the above hybrid encoder and decoder. Experiments indicate that Nana-HDR gives full play to the advantages of each component, such as strong text encoding ability of Transformer-based encoder, stateful decoding without being bothered by exposure bias and local information preference, and stable alignment provided by duration predictor. Due to these advantages, Nana-HDR achieves competitive performance in naturalness and robustness on two Mandarin corpora.