Reproduce “Dyads: Artist-Centric, AI-Generated Dance Duets” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zixuan Wang, Luis Zerkowski, Ilya Vidrin, Mariel Pettee
Dyads developed an AI method to generate choreographic dance duets, focusing on modeling the complex interactions between two dancers using a VAE-Transformer architecture with a custom velocity loss. This artist-centric project provides choreographers with a tool for creative exploration, yielding outputs that reflect artistic practice principles and spark discussions on partnering dynamics.
Key insights
- Modeling dance as a social practice, rather than just individual movement, is crucial for capturing the essence of many dance forms and human interaction.
- Incorporating artist input throughout the AI development lifecycle leads to tools that are not only technically sound but also aesthetically relevant and genuinely useful for artistic inquiry.
- Even physically implausible or unexpected AI-generated movements can serve as powerful prompts for artistic exploration and critical discussions about choreographic principles and limitations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.