NNSVS: A Neural Network-Based Singing Voice Synthesis Toolkit

TL;DR

NNSVS is an open-source toolkit for neural network-based singing voice synthesis, offering advanced features like multi-stream models and neural vocoders, with experimental results showing significant performance improvements.

Contribution

It introduces a comprehensive, feature-rich open-source singing voice synthesis toolkit that outperforms previous systems and provides extensive documentation and scripts.

Findings

Our best system significantly outperforms Sinsy.

NNSVS provides extensive features and documentation.

Experimental results validate the effectiveness of NNSVS.

Abstract

This paper describes the design of NNSVS, an open-source software for neural network-based singing voice synthesis research. NNSVS is inspired by Sinsy, an open-source pioneer in singing voice synthesis research, and provides many additional features such as multi-stream models, autoregressive fundamental frequency models, and neural vocoders. Furthermore, NNSVS provides extensive documentation and numerous scripts to build complete singing voice synthesis systems. Experimental results demonstrate that our best system significantly outperforms our reproduction of Sinsy and other baseline systems. The toolkit is available at https://github.com/nnsvs/nnsvs.