The Name-Free Gap: Policy-Aware Stylistic Control in Music Generation

2025-08-30

AI Summary

This study addresses the fine-grained style control challenge in music generation arising from policy-driven bans on artist names, introducing the concept of the “nameless gap” to demonstrate that name masking alone fails to prevent stylistic imitation. Methodologically, we employ large language models to generate human-readable, reproducible style descriptors as artist-name substitutes; build upon MusicGen-small; and integrate VGGish and CLAP embeddings to design distribution- and segment-level similarity analyses alongside a novel Minimum Distance Attribution (MDA) metric for quantifying prompt controllability. Experiments show that named prompts yield strongest control, yet nameless descriptors recover ~85% of style alignment performance; cross-artist transfer degrades alignment, confirming their capacity to encode artist-specific stylistic signals. Key contributions include: (1) the first systematic evaluation protocol for nameless style control, (2) the MDA attribution metric, and (3) a reproducible, compliance-aware paradigm balancing regulatory adherence with controllability.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Web data quality in the era of algorithmically-generated content
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization

Abstract

Text-to-music models capture broad attributes such as instrumentation or mood, but fine-grained stylistic control remains an open challenge. Existing stylization methods typically require retraining or specialized conditioning, which complicates reproducibility and limits policy compliance when artist names are restricted. We study whether lightweight, human-readable modifiers sampled from a large language model can provide a policy-robust alternative for stylistic control. Using MusicGen-small, we evaluate two artists: Billie Eilish (vocal pop) and Ludovico Einaudi (instrumental piano). For each artist, we use fifteen reference excerpts and evaluate matched seeds under three conditions: baseline prompts, artist-name prompts, and five descriptor sets. All prompts are generated using a large language model. Evaluation uses both VGGish and CLAP embeddings with distributional and per-clip similarity measures, including a new min-distance attribution metric. Results show that artist names are the strongest control signal across both artists, while name-free descriptors recover much of this effect. This highlights that existing safeguards such as the restriction of artist names in music generation prompts may not fully prevent style imitation. Cross-artist transfers reduce alignment, showing that descriptors encode targeted stylistic cues. We also present a descriptor table across ten contemporary artists to illustrate the breadth of the tokens. Together these findings define the name-free gap, the controllability difference between artist-name prompts and policy-compliant descriptors, shown through a reproducible evaluation protocol for prompt-level controllability.

Problem

• Achieving fine-grained stylistic control in text-to-music generation
• Addressing policy compliance when artist names are restricted
• Evaluating lightweight descriptors as alternative to artist-name conditioning

Innovation

• LLM-generated descriptors for style control
• Policy-robust alternative to artist names
• Reproducible evaluation protocol for controllability