Beyond Artifacts: Towards Generalizable Synthetic Song Detection via Music-Intrinsic Features

2026-06-15

AI Summary

Existing methods for synthetic singing detection rely on low-level artifacts or fixed-feature assumptions, limiting their generalization capability. This work proposes Sofia, a novel framework that systematically investigates, for the first time, the role of intrinsic musical characteristics—such as vocal qualities, audio effects, and global structural patterns—in generator-agnostic detection. Sofia employs a flexible mixture-of-experts (MoE) architecture, integrating specialized feature experts with an adaptive gating mechanism to enable efficient and robust feature fusion. Concurrently, we introduce MUSIC8K, a new benchmark dataset for evaluation. Experimental results demonstrate that Sofia achieves an F1 score on MUSIC8K-O that surpasses the strongest baseline by 18.5 percentage points, while also exhibiting exceptional robustness across diverse conditions.

Abstract

The rapid advancement of AI music generators highlights the urgent need for reliable Synthetic Song Detection (SSD). Existing SSD methods often rely on low-level artifacts or fixed feature assumptions, struggling to capture generator-agnostic cues. To address this, we propose Sofia (Synthetic-song detection framework via music features), a flexible framework that models music-intrinsic attributes via feature-specific experts and an adaptive Mixture-of-Experts (MoE) module. By configuring Sofia with representative Vocal, Audio-effect, Global structure features, and their combinations, we present their individual and complementary contributions. To comprehensively evaluate our framework, we further construct MUSIC8K, a challenging benchmark featuring lastest emerging generators and realistic audio perturbations. Experiments show that Sofia learns generator-agnostic representations from music-intrinsic features, improving the F1 score by 18.5 points over the strongest baseline on MUSIC8K-O while maintaining strong robustness.

Problem

• Synthetic Song Detection
• Generalizability
• Music-Intrinsic Features
• Generator-Agnostic
• AI Music Generators

Innovation

• music-intrinsic features
• Mixture-of-Experts
• generator-agnostic detection
• synthetic song detection
• MUSIC8K