Language-Codec: Bridging Discrete Codec Representations and Speech Language Models

TL;DR

Language-Codec introduces a novel approach to bridge discrete acoustic codecs with speech language models, improving generation quality and efficiency through innovative quantization and model design, validated by extensive evaluations.

Contribution

It proposes the Masked Channel Residual Vector Quantization (MCRVQ) and other enhancements to address gaps between codecs and speech models, advancing audio compression and downstream speech tasks.

Findings

Outperforms existing audio compression algorithms

Enhances downstream speech model efficiency

Validated through extensive evaluations

Abstract

In recent years, large language models have achieved significant success in generative tasks related to speech, audio, music, and other signal domains. A crucial element of these models is the discrete acoustic codecs, which serve as an intermediate representation replacing the mel-spectrogram. However, there exist several gaps between discrete codecs and downstream speech language models. Specifically, 1) Due to the reconstruction paradigm of the Codec model and the structure of residual vector quantization, the initial channel of the codebooks contains excessive information, making it challenging to directly generate acoustic tokens from weakly supervised signals such as text in downstream tasks. 2) numerous codebooks increases the burden on downstream speech language models. Consequently, leveraging the characteristics of speech language models, we propose Language-Codec. In the Language-Codec, we introduce a Masked Channel Residual Vector Quantization (MCRVQ) mechanism along with improved fourier transform structures and attention blocks, refined discriminator design to address the aforementioned gaps. We compare our method with competing audio compression algorithms and observe significant outperformance across extensive evaluations. Furthermore, we also validate the efficiency of the Language-Codec on downstream speech language models. The source code and pre-trained models can be accessed at https://github.com/jishengpeng/languagecodec .