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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Changhao, Chen, Jiahao, Xiang, Zhenghao, Yang, Zhixiong, Wang, Hanchen, Zhuang, Jiabao, Che, Xinmeng, Sun, Jiajun, Li, Hui, Cao, Yifei, Dou, Shihan, Zhang, Ming, Ye, Junjie, Ji, Tao, Gui, Tao, Zhang, Qi, Huang, Xuanjing
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911368027832320
author Jiang, Changhao
Chen, Jiahao
Xiang, Zhenghao
Yang, Zhixiong
Wang, Hanchen
Zhuang, Jiabao
Che, Xinmeng
Sun, Jiajun
Li, Hui
Cao, Yifei
Dou, Shihan
Zhang, Ming
Ye, Junjie
Ji, Tao
Gui, Tao
Zhang, Qi
Huang, Xuanjing
author_facet Jiang, Changhao
Chen, Jiahao
Xiang, Zhenghao
Yang, Zhixiong
Wang, Hanchen
Zhuang, Jiabao
Che, Xinmeng
Sun, Jiajun
Li, Hui
Cao, Yifei
Dou, Shihan
Zhang, Ming
Ye, Junjie
Ji, Tao
Gui, Tao
Zhang, Qi
Huang, Xuanjing
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03973
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Muse: Towards Reproducible Long-Form Song Generation with Fine-Grained Style Control
Jiang, Changhao
Chen, Jiahao
Xiang, Zhenghao
Yang, Zhixiong
Wang, Hanchen
Zhuang, Jiabao
Che, Xinmeng
Sun, Jiajun
Li, Hui
Cao, Yifei
Dou, Shihan
Zhang, Ming
Ye, Junjie
Ji, Tao
Gui, Tao
Zhang, Qi
Huang, Xuanjing
Sound
Computation and Language
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.
title Muse: Towards Reproducible Long-Form Song Generation with Fine-Grained Style Control
topic Sound
Computation and Language
url https://arxiv.org/abs/2601.03973