StrikeForge — Pose Preview

-- fps

▶

Loading runtime…

First-load downloads ~28 MB of models, cached after.

About This Preview

Pose detection runs entirely in your browser via TensorFlow.js MoveNet (SinglePose Lightning). No data leaves your device.

17 COCO keypoints: nose, eyes, ears, shoulders, elbows, wrists, hips, knees, ankles.

Analyze swing: records up to 4 s of keypoints (tap again to stop early), then a swing-phase detector runs in your browser — nothing is uploaded.

📷 Front / Back: switch between the selfie and rear camera. Rear is for filming a swing on a tripod / down-the-line.

● Capture session: tap to start, hit a bunch of swings, tap again to end. Unlike Analyze, this does upload the recorded clip (used to improve club & clubhead tracking). Opt-in — only what you record after tapping.

Swing Analysis

—

Coach’s read

v0 accuracy note: this detector was trained on broadcast golf footage using a different pose model than the one running in your browser, so expect rough edges — phase timings can be off, especially around Top and Impact, and on framings unlike TV coverage (selfie angle, partial body, non-golf motion). Treat the phases and tempo as a rough first read, not a measurement.

Handedness