R2-SVC: Towards Real-World Robust and Expressive Zero-shot Singing Voice Conversion

TL;DR

This work tackles the gap between idealized SVC training and real-world deployment by addressing noise, reverberation, and separation artifacts. It introduces R2-SVC, a three-pronged framework comprising simulation-based robustness augmentation, a singing-informed timbre/style extractor, and Neural Source-Filter (NSF) acoustic modeling to maintain semantic fidelity and expressive quality. Empirical results on both clean and noisy benchmarks demonstrate state-of-the-art performance, with clear gains in speaker similarity and naturalness and robust behavior under challenging conditions. The approach enhances the practicality of zero-shot SVC for real-world applications in music production, dubbing, and interactive voice systems.

Abstract

In real-world singing voice conversion (SVC) applications, environmental noise and the demand for expressive output pose significant challenges. Conventional methods, however, are typically designed without accounting for real deployment scenarios, as both training and inference usually rely on clean data. This mismatch hinders practical use, given the inevitable presence of diverse noise sources and artifacts from music separation. To tackle these issues, we propose R2-SVC, a robust and expressive SVC framework. First, we introduce simulation-based robustness enhancement through random fundamental frequency ($F_0$) perturbations and music separation artifact simulations (e.g., reverberation, echo), substantially improving performance under noisy conditions. Second, we enrich speaker representation using domain-specific singing data: alongside clean vocals, we incorporate DNSMOS-filtered separated vocals and public singing corpora, enabling the model to preserve speaker timbre while capturing singing style nuances. Third, we integrate the Neural Source-Filter (NSF) model to explicitly represent harmonic and noise components, enhancing the naturalness and controllability of converted singing. R2-SVC achieves state-of-the-art results on multiple SVC benchmarks under both clean and noisy conditions.

Figures (2)

• Figure 1: Architecture overview of R2-SVC.
• Figure 2: Mel-spectrogram visualizations comparing R2-SVC with Seed-VC and FreeSVC.