Research-grade motion analysis

Every joint.
Every frame.
Every rep.

32-point skeletal tracking in your browser. Real-time form analysis against elite references. An AI coach grounded in biomechanics — not vibes.

Open Motion Dashboard Read the paper
LIVE DASHBOARD — JOINT OVERLAY
L KNEE142°
R KNEE139°
L HIP76°
R HIP74°
TRUNK07°
L ANKLE94°
AI COACH
Valgus collapse at frame 247. Glute med engagement low. Hip external rotation −12°. Foot tripod unstable.
0.98
Vicon correlation
vs. marker-based gold standard
<10ms
Per-frame latency
On-device, any modern device
<2.4°
Mean angular error
vs. 12-camera marker capture
32
Joint vertices
Tracked per frame, per rep
THE PIPELINE

From a phone clip to a coach-grade report — in seconds.

01

Capture

Open the dashboard on any device with a camera. Frame yourself. No markers, no suit, no studio setup required.

02

Track

MediaPipe BlazePose runs on-device via WebGL. 32 joint vertices estimated per frame at sub-10ms. Nothing leaves your device.

03

Compare

STCF-DTW aligns your reps to an elite reference even when speed differs. Frame-pair matching with 8 angle definitions per pose.

04

Coach

An LLM grounded in your joint data writes a plain-English breakdown. Why you deviated, what pattern is behind it, and what to fix.

THE DASHBOARD

Research-grade, end-to-end. Without the lab.

Real-time inference

Sub-10ms per-frame latency on any modern laptop or phone. Live joint overlay as you move, not playback analysis after.

No markers. No suit.

32 vertices tracked from any webcam. Works at 30–240 fps. No special hardware, studio, or calibration required.

STCF-DTW analysis

Spatiotemporal Coherence Fourier Dynamic Time Warping aligns your movement to elite references accounting for speed differences.

AI form coach

A GPT-powered coach explains every deviation in plain English, grounded in your actual joint angles — not generic form tips.

Reference pose library

200+ elite-athlete reference poses across rowing, golf, squat, gait, tennis, and clinical rehab protocols. Annotated frame-by-frame.

Session history

Unlimited analyzed sessions. Joint-angle timelines, deviation heatmaps, and week-over-week trends. Track improvement with real numbers.

Your video never leaves your device. Pose extraction is 100% on-device. Only joint angle JSON (not video) syncs when you choose to save a session — encrypted, opt-in.
VALIDATION

Verified against the gold standard.

The pose engine is benchmarked against marker-based motion capture (Vicon) — the gold standard in biomechanics research.

0.98
Hip and knee flexion across squat, sprint, and gait benchmarks — tested in published methodology.
<2.4°
Average per-joint error against marker-based ground truth. Worst-case (occluded ankle) under 5°.
94%
Agreement with marker-based capture across 40 participants and six common rehabilitation movements.
IN PROGRESS

Two physical therapy clinics are running parallel-arm comparisons against marker-based gait studies. Results publish Q3 2026.

Lab-grade analysis without the lab.

Motion runs where you work — web dashboard, iPhone in the field, Motion³ for clinical precision. All three surfaces share the same analysis engine, so the number you see in the clinic is the same number the athlete sees at home.

Motion Dashboard
Web. Upload a clip, get joint-level analysis and AI coaching in seconds.
Motion for iPhone
iOS. Capture and review sessions on the go. HealthKit-integrated.
Motion³
Hardware. Precision capture camera for clinical and research settings.