MMGenre: Benchmarking Singing Voice Synthesis across Multiple Musical Genres

2026-07-08

AI Summary

This work addresses the limited cross-genre generalization of existing singing voice synthesis systems and the absence of a unified evaluation benchmark. We propose MMGenre, the first multi-genre singing synthesis benchmark encompassing ten major categories and twenty-six sub-genres, along with an automated score alignment pipeline to facilitate model diagnostics. Through acoustic feature analysis and separability assessment, we reveal that mainstream models exhibit highly similar acoustic characteristics across genres and insufficient stylistic distinctiveness. Furthermore, we demonstrate that lightweight genre-specific fine-tuning significantly enhances the stylistic expressiveness of synthesized singing voices. Our contributions establish a standardized evaluation framework and provide an effective optimization pathway for multi-style singing voice synthesis.

Abstract

Singing voice synthesis (SVS) has progressed rapidly, yet its ability to generalize across diverse musical genres remains underexplored. Existing benchmarks are heavily biased toward pop music, limiting systematic analysis of genre-dependent behavior. We introduce MMGenre, a benchmark for multi-genre SVS diagnosis, supported by an automatic pipeline for constructing genre-aligned music scores. MMGenre spans 10 major genres and 26 subgenres, enabling comprehensive analysis of genre-aware synthesis. Extensive evaluation of representative SVS models reveals limited genre discrimination: synthesized vocals across genres exhibit highly similar acoustic characteristics and weak separability. While zero-shot genre adaptation yields only marginal improvements, lightweight genre-specific continued training leads to substantial gains. MMGenre provides a standardized framework for multi-genre SVS evaluation and exposes critical challenges in achieving genre-aware singing voice synthesis.

Problem

• singing voice synthesis
• musical genres
• genre generalization
• benchmarking
• multi-genre evaluation

Innovation

• singing voice synthesis
• multi-genre benchmark
• genre-aware synthesis
• automatic score alignment
• lightweight adaptation