A controlled test.
A better body scan.
Guided phone capture for the height-scaled arm experiment. Compare single frames, steady holds and controlled elbow movement using the same evidence.
Set up before recording
Support the phone so it cannot move. Use barefoot standing on firm, level ground. Frame the crown, feet and the entire arm movement with margins. Keep the camera level and use close-fitting clothing that leaves the arm visible. Keep the page open and test the spoken cues before stepping away. Mount the phone physically in portrait orientation for portrait modes. Do not zoom or rotate the phone during an attempt. Do not force any movement.
No microphone, cloud inference, account or calibration object is required. Neutral standing height anchors a manual 2D diagnostic only; perspective is not corrected in this version.
Position the phone, then stand where your full body and arm movement will remain visible.
Instructions are time-based. This build does not automatically verify anatomy, pose or framing. A recording can complete while its pose is unusable.
Export sessions before clearing browser data. Aborted and interrupted attempts remain visible.
Manual diagnostic baseline
First choose a N0 frame and mark the crown and floor beneath the feet. Then mark shoulder, elbow and wrist on arm frames. These are human-picked image points, not certified anatomical landmarks. The scale is fixed to N0; moving in depth invalidates it.
Per-block projected measurements
Spread between frames is not independent repeatability. Bent-pose shoulder-to-wrist chords are intentionally different from extended-pose chords. Compare like poses only.
Attempt-to-attempt manual baseline comparison
A change in projected length mixes annotation, pose, depth and camera errors. It is not a validated anatomical difference or a formal accuracy pass.