TempoDance AI is a local-first dance coach that turns a tutorial into a source-synchronized pose reference, compares it with the learner's movement, and gives one correction at a time.

This demo shows:

a native-rate 30 FPS COCO-17 coach overlay; source audio with detected 114 BPM counts; upper-body, lower-body, and combined practice steps; local pose comparison and per-limb feedback; session memory that evaluates whether a coaching cue worked before adapting.

The default demo runs without a camera or cloud API. In camera mode, frames are processed by the local FastAPI pose service and are not persisted by application code.

Live demo: https://tempodance-ai.vercel.app Code: https://github.com/CarlKho-Minerva/tempodance-ai Reference tutorial: https://www.youtube.com/shorts/jrUsvBKemBU