How Is Dancing Scored?

What does AI dance scoring measure?

When you submit a clip, Danzu does two analyses in parallel. The music is processed with AI beat tracking to recover the song's rhythmic grid — where every beat and accent falls. Your body is processed with pose estimation, which tracks your joints frame by frame and converts your movement into precise motion data: joint angles, velocities, and accelerations over time.

Scoring happens where those two streams meet. Danzu measures how your movement relates to the music — when you accent, how continuously you ride the rhythm, how crisply your hits land, how big you move, how much new material you show, and how your intensity tracks the song's structure — and rates each of those qualities against the standards of the genre you picked.

The six axes of a dance score

One overall number can tell you that something is off, but never what. Danzu splits the score into six axes so the feedback is actionable:

• Timing — how precisely your accents land on the beat.
• Groove — how naturally your body rides the rhythm between the accents.
• Sharpness — how crisply your moves snap on the beat — hit versus smear.
• Amplitude — how big you move — how far your joints travel, full-out or marking.
• Variety — how broad your movement vocabulary is — new material versus looping.
• Dynamics — how your intensity tracks the song's contrast — chorus bigger than verse.

Each axis page explains what the axis means, how it is scored, and how to train it. Read together, the six scores are a diagnosis: accurate but stiff means train groove; big but mushy means train sharpness.

Why Danzu doesn't need a reference video

Most dance-feedback apps score you by comparing your body to a reference video, frame by frame — which means they can only judge how well you copy someone else's choreography. Danzu is reference-free: it scores your own movement against genre-aware standards, so freestyle, improvisation, and original choreography are all fully scoreable. Nothing to memorize, nothing to copy. Curious how that differs from other apps? See how Danzu compares.

How accurate is AI dance scoring?

Danzu's analysis is deterministic: the same video always produces the same score. That makes scores consistent and comparable over time — if your timing score rises from one week to the next, your timing got better, not the mood of an algorithm. The analysis also filters out camera shake and video noise, so it measures your body, not your phone.

Scores are AI estimates built for training and motivation. They are not competition adjudication, and they don't replace a teacher — they give you the honest, repeatable measurement a mirror can't.

How to raise your dance score

The loop is simple: film, read the six axes, train the lowest one, film again. Danzu's coaching tips are generated from your own clip, so every session points at the specific thing to fix next — and because analysis takes about 30 seconds, you can run the loop as many times per session as you have takes in you. Wondering where you'd land right now? Start with an honest look at your level.

Want the technical version?

This page covers what the six axes mean and how to read a score. How AI analyses and scores dancing goes a level down: the pose-estimation and beat-tracking pipeline, which movement features feed which axis, and an honest list of what the method cannot measure.