ESPnet-Codec: Comprehensive Training and Evaluation of Neural Codecs for Audio, Music, and Speech
Jiatong Shi, Jinchuan Tian, Yihan Wu, Jee-weon Jung, Jia Qi Yip, Yoshiki Masuyama, William Chen, Yuning Wu, Yuxun Tang, Massa Baali, Dareen Alharhi, Dong Zhang, Ruifan Deng, Tejes Srivastava, Haibin Wu, Alexander H. Liu, Bhiksha Raj, Qin Jin, Ruihua Song, Shinji Watanabe
TL;DR
This work introduces ESPnet-Codec, an open-source platform built on ESPnet to train and evaluate neural codecs across audio, music, and speech, complemented by VERSA, a versatile evaluation toolkit with intrusive, non-intrusive, and perceptual metrics. The authors demonstrate seamless integration with six ESPnet downstream tasks via a discretization interface and provide extensive experiments on LibriTTS and AMUSE datasets, showing that no single codec dominates across all metrics and that large-scale data can shift comparative performance. The study also highlights how codec-based representations affect downstream tasks such as ASR, TTS, SPK, SSE, SVS, and SSL, revealing both opportunities and challenges in codec-driven workflows. Overall, ESPnet-Codec and VERSA offer a standardized, comprehensive framework for fair comparisons, robust evaluation, and practical deployment of neural codecs in real-world speech and audio systems, with implications for efficient compression and downstream learning.
Abstract
Neural codecs have become crucial to recent speech and audio generation research. In addition to signal compression capabilities, discrete codecs have also been found to enhance downstream training efficiency and compatibility with autoregressive language models. However, as extensive downstream applications are investigated, challenges have arisen in ensuring fair comparisons across diverse applications. To address these issues, we present a new open-source platform ESPnet-Codec, which is built on ESPnet and focuses on neural codec training and evaluation. ESPnet-Codec offers various recipes in audio, music, and speech for training and evaluation using several widely adopted codec models. Together with ESPnet-Codec, we present VERSA, a standalone evaluation toolkit, which provides a comprehensive evaluation of codec performance over 20 audio evaluation metrics. Notably, we demonstrate that ESPnet-Codec can be integrated into six ESPnet tasks, supporting diverse applications.