Robust Singing Voice Transcription Serves Synthesis

TL;DR

This paper introduces ROSVOT, a robust note-level singing voice transcription model that improves accuracy and robustness for practical singing dataset annotation, enhancing Singing Voice Synthesis applications.

Contribution

The paper presents ROSVOT, the first robust AST model tailored for SVS, with a multi-scale framework and attention-based pitch decoder, enabling reliable transcription in real-world conditions.

Findings

Achieves state-of-the-art accuracy on clean and noisy data

Outperforms baseline SVS models when trained on enlarged datasets

Demonstrates practical applicability with real-world annotations

Abstract

Note-level Automatic Singing Voice Transcription (AST) converts singing recordings into note sequences, facilitating the automatic annotation of singing datasets for Singing Voice Synthesis (SVS) applications. Current AST methods, however, struggle with accuracy and robustness when used for practical annotation. This paper presents ROSVOT, the first robust AST model that serves SVS, incorporating a multi-scale framework that effectively captures coarse-grained note information and ensures fine-grained frame-level segmentation, coupled with an attention-based pitch decoder for reliable pitch prediction. We also established a comprehensive annotation-and-training pipeline for SVS to test the model in real-world settings. Experimental findings reveal that ROSVOT achieves state-of-the-art transcription accuracy with either clean or noisy inputs. Moreover, when trained on enlarged, automatically annotated datasets, the SVS model outperforms its baseline, affirming the capability for practical application. Audio samples are available at https://rosvot.github.io.