Muse: Towards Reproducible Long-Form Song Generation with Fine-Grained Style Control

TL;DR

Muse is an open-source system for long-form song generation that offers fine-grained style control, using a synthetic dataset and a simple training approach to achieve competitive quality and controllability.

Contribution

This work introduces a fully open-source, reproducible pipeline for long-form song generation with style control, including a synthetic dataset and a novel training method without complex architectural changes.

Findings

Muse achieves competitive performance on phoneme error rate.

Muse enables controllable segment-level music generation.

The system performs well despite modest data and model size.

Abstract

Recent commercial systems such as Suno demonstrate strong capabilities in long-form song generation, while academic research remains largely non-reproducible due to the lack of publicly available training data, hindering fair comparison and progress. To this end, we release a fully open-source system for long-form song generation with fine-grained style conditioning, including a licensed synthetic dataset, training and evaluation pipelines, and Muse, an easy-to-deploy song generation model. The dataset consists of 116k fully licensed synthetic songs with automatically generated lyrics and style descriptions paired with audio synthesized by SunoV5. We train Muse via single-stage supervised finetuning of a Qwen-based language model extended with discrete audio tokens using MuCodec, without task-specific losses, auxiliary objectives, or additional architectural components. Our evaluations find that although Muse is trained with a modest data scale and model size, it achieves competitive performance on phoneme error rate, text--music style similarity, and audio aesthetic quality, while enabling controllable segment-level generation across different musical structures. All data, model weights, and training and evaluation pipelines will be publicly released, paving the way for continued progress in controllable long-form song generation research. The project repository is available at https://github.com/yuhui1038/Muse.