Mocta-X · AI rowing video analysis — erg and on-water
Perfect Form.
Every Stroke
Film a piece on a phone. Rowing, running and golf come back graded stroke by stroke, with a coach that cites the frame — and any other movement can be compared against a reference.
Graded, explained, end‑to‑end.
Capture, tracking, grading, coaching and history — one pipeline from a phone clip to a session you can trend. Without the lab.
A live skeleton, no install.
8graded joint angles
No suit · no sensors · no studio- ROWING●
- RUNNING●
- GOLF●
- VOLLEYBALL
- + any movement vs a reference
Three graded, one for teams.
Progress, measured.
Every angle.
Drill any single frame.
Anywhere you train.
Three sports graded. Any movement compared.
The engine tracks 33 landmarks regardless of sport. What a coach-graded sport adds is phase segmentation, a rubric with published weights, target bands drawn from the literature, and a coach prompt that knows the sport.
Catch and finish angles, shin angle, knee compression, layback, leg drive, drive and recovery sequence. Erg, on-water, and a whole crew from one clip.
Rowing video analysis →Cadence, contact and swing time, vertical oscillation, pelvic drop, an Injury Load Index — targets fitted to your body, not a generic band.
Running form analysis →Phases and checkpoints through the swing, joint angles at each, and a reference archetype to compare against.
Court homography, player tracking and contact detection for team stats. Built for the coach to confirm, not full-auto.
Upload two clips, or a clip and a reference, and get eight joint angles aligned frame by frame with phase-aware DTW. No sport-specific rubric — the deltas are the verdict.
Elite plans upload a custom reference clip — a coach's demo, your best rep, a pro's swing — and every later session is compared against it.
See plansFrom a phone clip to a graded, explained session.
No black boxes. Each step is named and visible in the dashboard. If a number feels wrong, open the frame and see where it was measured.
Capture
Film side-on with any phone. The iPhone app has a capture wizard and trimmer; on the web you upload a clip or run the live camera. No markers, no suit, no studio.
Track
MediaPipe Pose places 33 landmarks on every frame and the angles are corrected for the video's aspect ratio. Realtime practice runs the model in your browser; uploaded clips are tracked on our servers.
Grade
A sport grader segments the movement into phases — catch, drive, finish, recovery for rowing — and scores each metric with continuous partial credit against a target band. Any other movement is aligned to a reference with phase-aware DTW.
Coach
A frontier model with tools queries the graded data — strokes, angles, consistency, your history — and writes a plain-language read that names the stroke and frame behind each claim.
Uploaded clips are graded on our servers. Here is what that means.
Footage is stored privately, reached only through time-limited links, encrypted in transit and at rest, and deleted with your account. Realtime practice never uploads video, and the iPhone app can run pose extraction on-device so only joint positions leave the phone. We never send raw video to an AI provider, and we do not run face recognition.
One product. Two surfaces.
The web dashboard is the flagship and where coaches work. The iPhone app is for the boathouse, the erg room and the road — and it shares every session with the web.
Film on the phone, review on the laptop — or the other way round. One account.
Measured,noteyeballed.
The parts of Mocta-X Dashboard that do the work — sport-specific graders, a coach with tools, and a pipeline that refuses to grade what it cannot see.
Graded stroke by stroke
Rowing, running and golf each have a coach-built rubric. Every stroke, stride and swing is segmented into phases and scored against target bands — with the weights published in the scoring guide.
A coach that investigates first
The AI coach runs on a frontier model with tools that query your strokes, angles and history. It shows the trail, then writes the read — and it cites the frame.
Nothing to install
Open app.mocta-x.com in any browser and upload a clip. Realtime practice runs the pose model in your browser; uploaded clips are graded on our servers and the result comes back in about the length of the clip.
33-point skeletal tracking
Markerless pose estimation on every frame from any phone or laptop camera — no suits, no sensors, no studio. Angles are corrected for the video's aspect ratio so a 25° lean reads as 25°.
Honest about what it can't see
Capture-quality gates flag a clip that is too far, too dark, too short or occluded before it is graded, and the grade is withheld when the clip can't support one.
History, progress and reports
Every session is saved and trended. Six-axis ability radar, a focus card, deep-report PDFs, and shareable links to hand a coach.
The real thing, not a mock-up.
One thing to work on, ranked by points.
Not a wall of numbers. The metric worth the most points comes first — your value, the target band, a cue, and the stroke to watch — beside a six-axis ability radar for the session.
- Ranked by
- Points at stake
- Radar
- 6 axes
Every angle, every frame.
Thirty-three landmarks tracked through the full stroke. Turn on the inspector, click a joint on the video, step frame by frame, and watch the athlete-position figure and the handle path move with it.
- Landmarks
- 33
- Hardware
- Phone cam
Legs, back, arms — in that order?
When each body part hits peak speed, drive and recovery separated, with the hand-off between segments scored. Click anywhere on the chart and the video jumps to that frame.
- Segments
- Legs · back · arms
- Checks
- Drive + recovery
A coach that reads the session before it speaks.
A frontier model with tools. It investigates your strokes, angles, consistency and history — the trail is shown — then writes a deep read that names the stroke and the frame behind every claim.
- Model
- Frontier, with tools
- Grounding
- Tool calls on your data
Pick a sport. Click analyze. Watch the dashboard run.
This is the live dashboard's image-analysis surface. Every keypoint, angle, and score you see was extracted offline by MediaPipe Pose — same code path the paid analysis runs. Pick a sport and watch the gauge, charts, 3D skeleton, and AI coach stream in.
/image-analysis
Compare against a reference
Reference frame on the left. Your attempt on the right. Per-joint scoring, aligned, with a coach's read.
Rowing demo images loaded. Click "Analyze poses" above to run the comparison.
/input
Reference image
Elite reference · saved
/input
Your image
f187 / 412
Pose data extracted offline by MediaPipe Pose · same code path the live dashboard uses.
Now do it for a whole clip.
Same flow, longer surface. Pick a session, hit analyze, and the studio loads — phase chips, deviation curve, score-tinted skeleton overlay, AI coach streaming the read.
/video-analysis
Compare a session against a reference
Phase-aware DTW alignment across the clip. Frame-level deviation. AI summary you can hand to a coach.
Rowing demo clip loaded. "Analyze video" runs the full studio pipeline.
Stored sessions — the live studio runs identical components on real uploads.
Measured on our own corpus. Published where we can cite it.
Single-camera pose estimation has known limits, and we would rather show you ours than quote a benchmark we have not run. These are the numbers we stand behind today; the methodology page has the rest, including what breaks.
Across the 57 stored rowing sessions, re-grading the same footage used to move a score by up to 24 points. After calibration it moves by 3 at worst, 0.6 on average.
Running reference ranges come from published regression equations fitted on 860 recreational runners and personalised to your sex, age, height and speed — not one band for everyone.
The personalised vertical-oscillation prediction was checked against an instrumented-treadmill report: 77.6 mm predicted, 77.2 mm measured.
The fast pose model is 4× quicker and costs about 8° on knee and elbow angles. Uploaded clips run the full model; the speed is not worth the error.
Internal measurements, not peer-reviewed results. The running cohort figure is from Orthop J Sports Med 2023 (n = 860). Rowing target bands draw on Kleshnev/BioRow, Concept2 and Egan-Shuttler et al. Full sources and limitations on the methodology page.
Built for every body in motion.
Start free. Upgrade when you need more.
Free
- 5 image · 2 video analyses a day
- 5 AI coach messages a day
- 30-day history
- Rowing, running and golf grading
Pro
Most popular- 50 image · 20 video analyses a day
- Streaming AI coach, 100 messages a day
- Unlimited history · PDF reports
- 500 MB storage · 2-minute clips
Elite
- Unlimited image and realtime analysis
- Unlimited AI coach · live voice coach
- 5 GB storage · unlimited clip length
- Custom references · priority support
Everything you'll want to ask.
Grouped by what people ask most. If yours isn't here, write to support and a real person answers — usually within a working day.
Still have questions? support@cam-lab.org
Open the dashboard. Run your first session.
No install, no credit card. The free plan grades clips every day — five images and two videos — and the AI coach is included.
Open Dashboard