Miguel O'Malley

GitHub

Autogeneration for DDR/ITG stepcharts using AI

DDCL

Dance Dance Revolution (DDR) and its modern counterpart, In the Groove, are popular rhythm games in which players step on arrows in time with music and are scored based on how well they match the timing of the arrows. The process of creating a DDR/ITG stepchart is challenging, requiring both musical knowledge and deep understanding of game mechanics.

In our work in Dance Dance ConvLSTM (DDCL), we explore the use of machine learning to generate stepcharts for DDR/ITG. We introduce a Convolutional LSTM (ConvLSTM) based encoder to learn beat-timing based motifs in music and translate them into playable stepcharts. Our model is trained on a dataset of existing stepcharts, learning to recognize the structure and timing of the arrows. Once trained, the model can autonomously generate new stepcharts that are both playable and challenging.

DDCLarchitecture

The results of our work show that machine learning can be used to create playable stepcharts for DDR/ITG, opening up new possibilities for game design and music interaction. We believe that our ConvLSTM based approach can be extended to other tasks in music information retrieval (MIR), such as providing encoder structure for transformer based architectures, as we consider recognition of the beat to be a fundamental differentiator between general audio processing tasks and musical ones.