FreeCodec: A disentangled neural speech codec with fewer tokens

TL;DR

FreeCodec introduces a novel neural speech codec that effectively disentangles speech components into timbre, prosody, and content, achieving superior reconstruction and disentanglement with fewer tokens.

Contribution

It proposes a new encoding framework that decomposes speech into distinct components, improving coding efficiency and performance over residual vector quantization methods.

Findings

Outperforms existing methods in reconstruction quality.

Achieves state-of-the-art disentanglement performance.

Demonstrates effectiveness with fewer tokens.

Abstract

Neural speech codecs have gained great attention for their outstanding reconstruction with discrete token representations. It is a crucial component in generative tasks such as speech coding and large language models (LLM). However, most works based on residual vector quantization perform worse with fewer tokens due to low coding efficiency for modeling complex coupled information. In this paper, we propose a neural speech codec named FreeCodec which employs a more effective encoding framework by decomposing intrinsic properties of speech into different components: 1) a global vector is extracted as the timbre information, 2) a prosody encoder with a long stride level is used to model the prosody information, 3) the content information is from a content encoder. Using different training strategies, FreeCodec achieves state-of-the-art performance in reconstruction and disentanglement scenarios. Results from subjective and objective experiments demonstrate that our framework outperforms existing methods.