Anonymous authors (paper currently under review)
MusGU+ (Music-Generative Usable+ AI) is a musician-centered evaluation framework designed to assess how generative music models can be adapted, used, and controlled in real-world creative contexts.
🔍 Explore the MusGU+ discovery tool
MusGU+ evaluates generative music models along three complementary dimensions, each framed around a core practical question:
- Adaptability — Can I realistically adapt this model to my own data?
- Usability — Can I access, run, and integrate this model into my music-making workflow?
- Controllability — Can I guide the model in musically meaningful and interpretable ways?
Each dimension is composed of multiple criteria (e.g., hardware requirements, interface availability, conditioning inputs, control parameters) and evaluated on a three-level scale depending on the degree of support provided: fully, partially, not.
📖 Read the detailed evaluation criteria.
Rather than presenting a fixed leaderboard, MusGU+ is designed as an interactive discovery tool. It allows musicians to explore, filter, and compare generative music models based on specific criteria and tags, highlighting differences in adaptability, usability, and controllability. This supports early-stage exploration and informed selection of models that may fit particular creative practices or workflow needs.
MusGU+ builds on insights from the MusGO framework. MusGO (Music-Generative Open AI) is an openness-focused evaluation framework for music-generative AI. While MusGO assesses transparency and responsible research practices, MusGU+ supports informed selection and practical adoption of generative music models by musicians.