DanzaKuduro

Inspiration

As dancers ourselves, we know how tough it is to create choreography and then step back to judge it. Are we repeating moves too much? Is the routine too fast to stay on beat? We wanted a tool built by dancers for dancers. Something that not only refines choreography but also helps us grow by showing exactly where we’re hitting the mark and where we’re not.

What it does

DanzaKuduro lets users upload dance videos, have them analyzed by AI, and receive instant feedback report. The system highlights where moves match the beat, where timing slips, and when choreography repeats. Performances are organized by song, so dancers can compare takes, track progress, and polish routines.

How we built it

We built the frontend in React with a clean interface for organizing songs and videos. For motion analysis, we experimented with different AI models and integrated MediaPipe. This works with a backend in Python. Finally, we used gpt-oss (20b) to use this data to create a valuable feedback report for the user.

Challenges we ran into

One major challenge was finding a motion-tracking model accurate enough for dance. YOLO and similar object-detection models struggled with accuracy, especially when multiple dancers were in frame. Ultimately, MediaPipe delivered good results for solo performances, but we plan to expand into better multi-person tracking solutions for group choreographies.

Accomplishments that we're proud of

1. Designing a working AI pipeline that can give meaningful feedback.
2. Building a clean, easy-to-use interface where videos can be uploaded and organized by song, making analysis more structured.
3. Creating something that reflects our own needs as dancers and can help others too.

What we learned

We learned how to connect AI motion tracking with a clean frontend, and what feedback really matters to dancers, like timing, speed, and repetition. We also saw the difficulty in multi-person tracking is, and how much small design choices improve the user experience.

What's next for DanzaKuduro

Next, we want to expand the report with new features, like:

1. A difficulty meter for choreography.
2. Analysis of a dancer’s energy levels.
3. Feedback reports for group performances, when multiple people are in a single frame.

Built With

• gpt-oss
• javascript
• python
• react
• vite