Abstract
We introduce a model that generates content for "osu! mania", a key-based rhythm game. We adapt various concepts from onset detection and beat tracking to construct a model that can generate "beatmaps" for any audio conditioned on a given difficulty value.