HiFi-Codec: Group-residual Vector quantization for High Fidelity Audio Codec

By Dongchao Yang, Songxiang Liu, Rongjie Huang, Jinchuan Tian, Chao Weng, Yuexian Zou. 2023

231 citations recorded by the source corpus.

Audio codec models are widely used in audio communication as a crucial technique for compressing audio into discrete representations. Nowadays, audio codec models are increasingly utilized in generation fields as intermediate representations. For instance, AudioLM is an audio generation model that uses the discrete representation of SoundStream as a training target, while VALL-E employs the Encodec model as an intermediate feature to aid TTS tasks. Despite their usefulness, two challenges persist: (1) training these audio codec models can be difficult due to the lack of publicly available training processes and the need for large-scale data and GPUs; (2) achieving good reconstruction performance requires many codebooks, which increases the burden on generation models. In this study, we propose a group-residual vector quantization (GRVQ) technique and use it to develop a novel \textbf{Hi}gh \textbf{Fi}delity Audio Codec model, HiFi-Codec, which only requires 4 codebooks. We train all the models using publicly available TTS data such as LibriTTS, VCTK, AISHELL, and more, with a total duration of over 1000 hours, using 8 GPUs. Our experimental results show that HiFi-Codec outperforms Encodec in terms of reconstruction performance despite requiring only 4 codebooks. To facilitate research in audio codec and generation, we introduce AcademiCodec, the first open-source audio codec toolkit that offers training codes and pre-trained models for Encodec, SoundStream, and HiFi-Codec. Code and pre-trained model can be found on: \href{https://github.com/yangdongchao/AcademiCodec}{https://github.com/yangdongchao/AcademiCodec}

Subfields: Artificial Intelligence, Model Architectures, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xu Tan
2. Jinyu Li
3. Dong Yu
4. Zhou Zhao
5. Tao Qin
6. Tie-Yan Liu
7. Yi Ren
8. Sheng Zhao

In Model Architectures

1. Xu Tan
2. Yi Ren
3. Zhou Zhao
4. Jinglin Liu
5. Sheng Zhao
6. Jinyu Li
7. Rongjie Huang
8. Yanqing Liu

In Speech & Audio

1. Xu Tan
2. Jinyu Li
3. Yi Ren
4. Zhou Zhao
5. Rongjie Huang
6. Jinglin Liu
7. Sheng Zhao
8. Dong Yu

In Signal Processing

1. Jinyu Li
2. Yi Ren
3. Xu Tan
4. Zhou Zhao
5. Rongjie Huang
6. Jinglin Liu
7. Dong Yu
8. Sheng Zhao

Citations in the corpus

• Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers (2023, neutral): “The new codecs include Vocos [62], SpeechTokenizer [63], AudioDec [64], AcademiCodec [65], Descript-audio-codec (DAC) [66], Fun-Codec [67], and RepCodec [68].”
• VALL-E 2: Neural Codec Language Models are Human Parity Zero-Shot Text to Speech Synthesizers (2024, background): “These include Vocos [Siuzdak, 2023], SpeechTokenizer [Zhang et al., 2023a], AudioDec [Wu et al., 2023], AcademiCodec [Yang et al., 2023a], Descript-audio-codec (DAC) [Kumar et al., 2024], FunCodec [Du et al., 2024b], and RepCodec [Huang et al., 2023b].”
• NaturalSpeech 3: Zero-Shot Speech Synthesis with Factorized Codec and Diffusion Models (2024, methodology): “We select the following open-source codec models as baselines: EnCodec [15] 12 , HiFi-Codec [65] 13 , and Descript-Audio-Codec (DAC) [32] 14 .”
• InstructTTS: Modelling Expressive TTS in Discrete Latent Space With Natural Language Style Prompt (2023, methodology): “The GRVQ-based audio codec model is the same as the one introduced in HiFi-Codec [24] and we use the officially open-source implementation 3 .”
• Make-An-Audio: Text-To-Audio Generation with Prompt-Enhanced Diffusion Models (2023, background): “, 2021) and HiFi-Codec (Yang et al., 2023) present a hierarchical architecture for high-level representations that carry semantic information.”
• Make-A-Voice: Unified Voice Synthesis With Discrete Representation (2023, methodology): “[43] propose a group-residual vector quantization (GRVQ) technique and use it to develop a high-fidelity audio codec model named HiFi-Codec.”
• Enhancing the Stability of LLM-based Speech Generation Systems through Self-Supervised Representations (2024, methodology): “As the tokens of an MLM are already highly correlated, because they can be predicted from other tokens in the same sequence, the causal language modeling task performed in LLMs is expected to be simpler when predicting these tokens than when predicting tokens with no enforced correlation, such as th”
• HierSpeech++: Bridging the Gap Between Semantic and Acoustic Representation of Speech by Hierarchical Variational Inference for Zero-Shot Speech Synthesis (2023, unknown): “Although all the above models have shown powerful adaptation performance, they have a slow inference speed for their iterative generation manner.”
• F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching (2024, unknown): “…impressive zero-shot performance as they perform implicit duration modeling and can leverage diverse sampling strategies, non-autoregressive (NAR) models benefit from fast inference through parallel processing, and effectively balance synthesis quality and Figure 1: An overview of F5-TTS…”
• SpeechTokenizer: Unified Speech Tokenizer for Speech Large Language Models (2023, methodology): “Discriminator We use the same dicriminators as HiFi-Codec Yang et al. (2023) that consist of three discriminators: A multi-scale STFT-based (MS-STFT) discriminator; a multi-period discriminator (MPD) and a multi-scale discrinator (MSD).”