SingVERSE: A Diverse, Real-World Benchmark for Singing Voice Enhancement

2025-09-25

AI Summary

Current singing voice enhancement research is hindered by the lack of real-world evaluation data. To address this, we introduce SingVERSE—the first benchmark specifically designed for singing voice enhancement under realistic acoustic conditions—featuring diverse recording scenarios and high-fidelity paired clean/distorted recordings. Leveraging SingVERSE, we conduct a systematic evaluation of state-of-the-art speech enhancement models, revealing for the first time an inherent trade-off between perceptual quality and intelligibility in singing tasks. We further validate the critical role of in-domain fine-tuning and propose a singing-aware data adaptation strategy. Experiments demonstrate that our approach significantly improves enhancement performance while preserving vocal naturalness. SingVERSE establishes a standardized, reproducible evaluation framework and provides principled optimization guidelines for singing voice enhancement research.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Natural Language Processing: Speech
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

This paper presents a benchmark for singing voice enhancement. The development of singing voice enhancement is limited by the lack of realistic evaluation data. To address this gap, this paper introduces SingVERSE, the first real-world benchmark for singing voice enhancement, covering diverse acoustic scenarios and providing paired, studio-quality clean references. Leveraging SingVERSE, we conduct a comprehensive evaluation of state-of-the-art models and uncover a consistent trade-off between perceptual quality and intelligibility. Finally, we show that training on in-domain singing data substantially improves enhancement performance without degrading speech capabilities, establishing a simple yet effective path forward. This work offers the community a foundational benchmark together with critical insights to guide future advances in this underexplored domain. Demopage: https://singverse.github.io

Problem

• Addresses the lack of realistic evaluation data for singing voice enhancement
• Introduces the first real-world benchmark covering diverse acoustic scenarios
• Investigates the trade-off between perceptual quality and intelligibility in models

Innovation

• Created real-world singing benchmark with clean references
• Evaluated trade-off between perceptual quality and intelligibility
• Showed in-domain training improves singing enhancement performance