HAM-TTS: Hierarchical Acoustic Modeling for Token-Based Zero-Shot Text-to-Speech with Model and Data Scaling

TL;DR

HAM-TTS introduces a hierarchical acoustic modeling approach with extensive data augmentation and model scaling, significantly improving zero-shot TTS speech naturalness, pronunciation accuracy, and style-timbre consistency.

Contribution

The paper presents a novel hierarchical acoustic model with data augmentation and large-scale training, advancing zero-shot TTS quality and diversity beyond existing models.

Findings

Outperforms VALL-E in pronunciation accuracy and style preservation.

Achieves high speech diversity with consistent timbre across generated samples.

Scales data and model size to 650k hours and 0.8B parameters, respectively.

Abstract

Token-based text-to-speech (TTS) models have emerged as a promising avenue for generating natural and realistic speech, yet they grapple with low pronunciation accuracy, speaking style and timbre inconsistency, and a substantial need for diverse training data. In response, we introduce a novel hierarchical acoustic modeling approach complemented by a tailored data augmentation strategy and train it on the combination of real and synthetic data, scaling the data size up to 650k hours, leading to the zero-shot TTS model with 0.8B parameters. Specifically, our method incorporates a latent variable sequence containing supplementary acoustic information based on refined self-supervised learning (SSL) discrete units into the TTS model by a predictor. This significantly mitigates pronunciation errors and style mutations in synthesized speech. During training, we strategically replace and duplicate segments of the data to enhance timbre uniformity. Moreover, a pretrained few-shot voice conversion model is utilized to generate a plethora of voices with identical content yet varied timbres. This facilitates the explicit learning of utterance-level one-to-many mappings, enriching speech diversity and also ensuring consistency in timbre. Comparative experiments (Demo page: https://anonymous.4open.science/w/ham-tts/)demonstrate our model's superiority over VALL-E in pronunciation precision and maintaining speaking style, as well as timbre continuity.