The foundation everything else is built on
Markerless pose tracking
Thirty-three body points, from a video you already have
Upload one clip filmed on a phone and get the position of thirty-three body points in every frame. No markers, no suit, no capture volume, nothing to calibrate.
- Points tracked
- 33 per frame
- Equipment
- One camera
- Longest clip
- 2 minutes
- Upload limit
- 500 MB
What you upload, and what comes back
A single video is the whole setup. The file is compressed in your browser before it is sent, so a long clip off a modern phone does not have to travel at full size, and pose estimation then runs on it frame by frame.
What comes back is a position for each of thirty-three body points in every frame — head, shoulders, elbows, wrists and hands, hips, knees, ankles, heels and toes — plus a rendered video with the skeleton drawn over the movement so you can see what the model actually tracked rather than taking it on trust.
What it can and cannot see
This measures what one camera can see. Angles are computed in two dimensions from the camera's viewpoint, so a joint rotating out of the filmed plane will read short, and occlusion, loose clothing, low frame rates and steep camera angles all cost accuracy. Filming square-on to the plane you care about, with the whole body in frame, is the single biggest thing under your control.
We have not published a validation study against marker-based motion capture. Treat the numbers as a repeatable measure of what is in the video rather than as a lab-grade ground truth, and read the skeleton overlay before you read the figures.