Mapperatorinator

Why we included this project

osu! mappers who want to generate or edit beatmaps will get real use out of this framework, which covers all four gamemodes and can also assist with modding existing maps. The pipeline runs a diffusion model with a transformer-based approach on spectrogram input, and you can shape the output with parameters such as difficulty star rating, mapper style, year, and style descriptors. Generation can be driven through a web UI, an interactive CLI, or command-line inference, and the release ships fine-tuned models for specific gamemodes along with an automatic model picker. The project is upfront about its compute-heavy training history and asks users to disclose AI use in their maps, a practical point for anyone publishing generated content. For mappers, it is a good way to prototype ideas, fill in sections, or try styles they would not normally attempt by hand.