DanceDuo: AI dance generator with user performance comparison
Pose estimation meets AI choreography in DanceDuo for interactive dance practice.
DanceDuo leverages diffusion models to create AI-choreographed dance sequences that are tightly synchronized with a variety of music genres. The system allows users to choose from multiple music tracks and humanoid avatars, then generates a full dance routine. What sets DanceDuo apart is its integration of human pose estimation: users can import their own dance videos and see a side-by-side comparison of their movements against the AI-generated choreography. This feedback loop enables precise self-assessment and improvement, blending generative AI with real-world human performance.
Presented at SOICT 2024, the platform underwent a comprehensive user study that highlighted the intuitive design and the popularity of the comparison feature. Participants found the interface easy to navigate, and the ability to visually benchmark their own dance against AI outputs was particularly praised. For professionals and hobbyists alike, DanceDuo offers a novel way to practice choreography, receive instant visual feedback, and explore creative dance ideas without requiring a partner or instructor. It represents a meaningful step in applying AI to embodied creativity and human-computer interaction.
- DanceDuo uses diffusion models to generate dance sequences synchronized with multiple music genres.
- Integrates human pose estimation for side-by-side comparison of user performances with AI-generated choreography.
- User study found the interface intuitive, with high praise for the dance comparison feature.
Why It Matters
Enables accessible AI-assisted dance practice and choreography for professionals and hobbyists alike.