Investigating Neural Audio Codecs for Speech Language Model-Based Speech Generation

TL;DR

This paper systematically studies how neural audio codec tokens influence speech generation in speech language models, revealing that high-quality codecs are essential for naturalness but not necessarily for intelligibility.

Contribution

It provides a comparative analysis of neural codec models within SLM frameworks, highlighting key factors affecting speech quality and intelligibility.

Findings

Better codec reconstruction does not always improve speech generation.

High-quality decoders are crucial for natural speech production.

Speech intelligibility depends more on quantization mechanisms.

Abstract

Neural audio codec tokens serve as the fundamental building blocks for speech language model (SLM)-based speech generation. However, there is no systematic understanding on how the codec system affects the speech generation performance of the SLM. In this work, we examine codec tokens within SLM framework for speech generation to provide insights for effective codec design. We retrain existing high-performing neural codec models on the same data set and loss functions to compare their performance in a uniform setting. We integrate codec tokens into two SLM systems: masked-based parallel speech generation system and an auto-regressive (AR) plus non-auto-regressive (NAR) model-based system. Our findings indicate that better speech reconstruction in codec systems does not guarantee improved speech generation in SLM. A high-quality codec decoder is crucial for natural speech production in SLM, while speech intelligibility depends more on quantization mechanism.