Open Arsenal

Song-to-Dance Intelligence

MotionNet is an open-source AI system designed to bridge music and movement by generating real-time ballroom dance choreography from any song. At its core, it analyzes audio input—detecting rhythm, tempo, structure, and emotional tone—to determine the most appropriate dance style. Once identified, it generates a synchronized virtual dancing couple that performs the choreography in real time, allowing users to visually learn and follow each movement as the music plays.

The platform is built around step-by-step instructional learning. Instead of only showing a finished routine, MotionNet breaks down each sequence into clear, timed actions, helping users understand posture, footwork, and transitions. This makes it useful for beginners learning ballroom fundamentals as well as advanced dancers refining timing and technique. The 3D visualization layer ensures that movements are intuitive and closely aligned with the music’s beat structure.

A key feature of MotionNet is its modular AI architecture. The system is divided into independent components such as music intelligence, choreography generation, motion rendering, streaming integration, and learning modules. Each module can be independently improved, replaced, or extended without disrupting the rest of the system. This design allows researchers and contributors to iterate rapidly and experiment with new approaches to dance generation and motion modeling.

MotionNet also includes a continuous learning pipeline driven by community contributions. Users can upload dance videos, which are analyzed to improve choreography accuracy, expand the system’s dance knowledge base, and refine motion realism over time. Combined with its streaming integration capabilities and open-source foundation under AGPL 3.0+, MotionNet is designed to evolve collaboratively into a living system for music-driven movement intelligence.