StrideLab

Pose-based running gait analysis. Upload or record a short clip and it is decoded, tracked and measured entirely in this browser — 40 biomechanical measurements, each with the confidence interval that goes with it. The video never leaves your device, because there is nowhere for it to go.

1 About you

Standing height required cm

Your height is what converts pixels into centimeters. The app measures your thigh and shin in the image and scales from there, so every distance below — stride length, bounce, how far ahead your foot lands — is only as accurate as this number.

Body mass unlocks stiffness kg Sex optional

Mass is what the spring-mass model needs to estimate vertical and leg stiffness — two of the four variables with a significant pooled association with running economy. Without it those two are not reported, because a population average would be inventing the answer. Sex is used only to pick a reference band where the literature gives one. Both stay on this device.

2 The clip

Camera view Surface Treadmill speed strongly recommended m/s

On a treadmill you do not move through the frame, so there is no displacement to measure. Without the belt speed there is no step length, no running speed, and no way to pick the right cadence band — a cadence target that ignores speed is folklore.

Drop a clip here or

3 How to film it

Shoot at 120 or 240 fps if the phone offers it. At 30 fps one frame is 33 ms and ground contact is only about 230 ms, so a third of the measurement is quantization. Below 60 fps this app refuses to report contact time at all.
Camera square to the running path, at hip height, on something that does not move. An angled camera measures angles out of the plane they happen in.
Fill 60–80% of the frame height with the runner, whole body, feet included.
Side on for timing, stride and sagittal angles. From behind for pelvic drop and crossover. They measure different things and neither substitutes for the other.
Three seconds minimum, six is better. The first and last stride are always discarded, and the numbers are aggregates over what is left.
Plain background, even light. Motion blur on a fast-moving foot is what breaks the foot landmarks, and the foot landmarks are what most of this depends on.

4 Engine settings

Filter cutoff

Zero-phase, forward and backward, with the dual-pass cutoff correction applied so the effective cutoff is the one shown. 12 Hz because impact transients and fast distal-segment motion in running live above the 6 Hz used for walking.

Use the GPU where available Run inference in a worker Keep the video with this analysis

Checking storage…

Start with a clip

Drop in a few seconds of running, or pick one of the two demos: a synthetic runner whose true numbers are known, or a real phone clip filmed on a treadmill.

Scores and key measurements appear here once a clip has been analyzed.

Johann Cardenas

StrideLab

Start with a clip

That did not work

Metric