How the analysis worksWhat the coach looks at — and what it's built on

1. You upload a short clip and pick your board type.
2. An AI model (Google Gemini) watches the ride frame by frame — not a single thumbnail, but the actual motion of your pop-up, trim, and turns.
3. It returns a structured coaching report: what's working, what to fix, the one priority, and a drill — plus still-frame captures at the key moments.
4. Your video is deleted immediately after analysis, from our server and from the AI provider. Only the text report is kept (see Privacy Policy).

The coach is told to describe what it actually sees in your ride first, and only then put it into coaching language. It is explicitly instructed not to invent movements the clip doesn't show, and to describe direction ("more bent", "a touch late") rather than guess exact numbers when something isn't clearly visible. The goal is honest feedback, not impressive-sounding filler.

Surfing isn't one sport. A classic longboard noseride and a shortboard snap are judged by completely different standards. So each board type has its own coaching reference:

• Classic longboard — noseriding, cross-step, flow
• Performance longboard — rail carving, power
• Midlength — trimming and basic turns, the in-between
• Shortboard — pop-up, bottom turn, snap
• Fish — loose rail-to-rail on small, soft waves
• Land training (new) — pop-up rehearsal or surf skate, judged on body mechanics with no wave to read

Crucially, this reference is used as a language guide, not a scorecard. The coach doesn't grade you against a checklist — it borrows the right words to explain why a movement matters, and it respects each board's nature (a fish's loose slide isn't a flaw by shortboard standards). For a land-training clip the coach first works out whether it's a pop-up rehearsal or a surf skate session, then applies only the matching mechanics — no fabricated "wave conditions" for a living-room clip.

To keep the coaching honest, we studied real pro surfing footage. Using pose estimation, we measured how professionals hold the key positions — the depth of a bottom-turn compression, the upright balance of a noseride, how front-side and back-side turns differ — across many riders.

Those findings are woven in as approximate reference ranges and directional tendencies— context to recognize good technique, not exact targets you must hit. Where the data couldn't reliably show something, we deliberately left it out rather than guess. (2D video can't measure precise joint angles, and we won't pretend otherwise.)

For land training, the references come from coaching practice — the same body mechanics (hand position, both-feet timing, knee bend, upper-body lead, rail-pressure rolling) that pros translate into the water — rather than from pose-extracted footage.

• It's an AI read — approximate, and a great human coach who knows you will still catch things it misses.
• Quality in, quality out: a clear, side-on clip with your whole body in frame gives a far sharper analysis than a distant, shaky one.
• It's built to be genuinely useful for self-coached surfers — a fast, objective second pair of eyes — not to replace time in the water.