MusGO Framework: Assessing Openness in Music-Generative AI

This website builds on the paper MusGO: A Community-Driven Framework for Assessing Openness in Music-Generative AI, authored by Roser Batlle-Roca, Laura Ibáñez-Martínez, Xavier Serra, Emilia Gómez, and Martín Rocamora, and published in the Proceedings of the 26th International Society for Music Information Retrieval Conference (ISMIR 2025). It serves not only as a companion to the publication, but also as a living resource, which is continuously updated and shaped by contributions from the community.

JAM

Singapore University of Technology and Design and Lamda Labs · 2025

A tiny flow-based song generator with fine-grained controllability and aesthetic alignment.

Application types: Text-to-songLyrics-to-song

Architecture:Latent diffusionRectified flowHybrid LLM

Essential categories

Category	Status	Notes
Source code	✔︎ Open	Source code is available in the GitHub repository, including data processing, training and inference code, as well as model architecture.
Training data	✘ Closed	Training data is not available nor properly described. In the research paper, they only indicate that they rely on a ~1 million song dataset, resulting into ~54k hours of audion. No information on specific data sources, acquisition methods, and data licensing is given. See "Dataset Setup" in Section 4.1. of the pre-print.
Model weights	✔︎ Open	Pre-trained version of the model is available in Hugging Face.
Code documentation	✔︎ Open	Codebase is documented, including high-level instructions and configuration files.
Training procedure	✔︎ Open	Training procedure is properly described, including hardware requirements and confirgurations (see Section 4.2).
Evaluation procedure	✔︎ Open	The evaluation procedure is properly described, including evaluation data, metrics, and implementation details (see Section 4.3). With the aim to standarize evaluation across models, they propose JAME: a genre-driverse evaluation dataset for full-track generation. This dataset is publicly available in Hugging Face (https://huggingface.co/datasets/declare-lab/JAME).
Research paper	~ Partial	Pre-print is available. No peer-reviewed version of the article has been found.
Licensing	~ Partial	JAM is released under the Project Jamify License and the Stability AI Community License Agreement, which allow for non-commercial use and research purposes with certain restrictions.

Desirable categories

Category	Status	Notes
Model card	⭐ Included	Model card available in Hugging Face.
Datasheet	∅ Not included	Dataset card for the evaluation dataset (JAME) is available in Hugging Face. However, no datasheet is provided for the training data used to train JAM.
Package	∅ Not included
User-oriented application	⭐ Included	There is a user-oriented application on Hugging Face Spaces, but it is currently not available. Nonetheless, all the files to run the app are provided in the HuggingFace repository. See https://huggingface.co/spaces/declare-lab/JAM/tree/main, and https://huggingface.co/spaces/declare-lab/JAM/blob/main/app.py.
Supplementary material page	⭐ Included	A supplementary material page is available showcasing the use of JAM.

Raw YAML file with complete evaluation.