Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers

TL;DR

This paper presents Vall-E, a neural codec language model for zero-shot text-to-speech synthesis that leverages large-scale training data and in-context learning to produce natural, personalized speech with minimal prompts.

Contribution

The paper introduces Vall-E, a novel neural codec language model that achieves high-quality zero-shot TTS by treating synthesis as a conditional language modeling task and scaling up training data.

Findings

Vall-E outperforms state-of-the-art zero-shot TTS systems in naturalness and speaker similarity.

It can synthesize personalized speech using only a 3-second speaker prompt.

Vall-E preserves speaker emotion and acoustic environment in synthesis.

Abstract

We introduce a language modeling approach for text to speech synthesis (TTS). Specifically, we train a neural codec language model (called Vall-E) using discrete codes derived from an off-the-shelf neural audio codec model, and regard TTS as a conditional language modeling task rather than continuous signal regression as in previous work. During the pre-training stage, we scale up the TTS training data to 60K hours of English speech which is hundreds of times larger than existing systems. Vall-E emerges in-context learning capabilities and can be used to synthesize high-quality personalized speech with only a 3-second enrolled recording of an unseen speaker as an acoustic prompt. Experiment results show that Vall-E significantly outperforms the state-of-the-art zero-shot TTS system in terms of speech naturalness and speaker similarity. In addition, we find Vall-E could preserve the speaker's emotion and acoustic environment of the acoustic prompt in synthesis. See https://aka.ms/valle for demos of our work.