Live Preview
See the invisible.
Nine working windows from the product suite: a movement search filling its results grid, a pressing pattern recognized as it forms, one player’s risk drift entering its threshold band, a 22-player readiness board, a running-gait check, a rehab trace descending into its baseline band, a run forecast 2.5 seconds ahead, a movement-signature shortlist, and a knowledgebase that answers with evidence. Everything is generated in your browser — no real match data is streamed here.
Window 01
Movement search
One query clip loops on the left; the grid fills with its closest recurrences from a 38-match season and re-sorts as better matches arrive. The capability behind it: Movement Search at Scale.
Query clip
Cut inside → sprint burst
Closest matches · similarity
- 0.96
- 0.94
- 0.91
- 0.89
- 0.87
- 0.85
- 0.82
- 0.80
Window 02
Tactical patterns
Five pressing players converge on the left half-space until the pattern is recognized and labeled with a confidence value. The capability behind it: Tactical Fingerprinting.
PATTERN RECOGNIZED
High press — left half-space
confidence 0.91
Window 03
Injury-risk drift
One player’s sessions drift away from his own baseline across twelve weeks; when the drift holds inside the threshold band, a single alert fires. The research program behind it: Injury Risk Signals.
Stride asymmetry — current vs. own baseline
- Own baseline
- Current sessions
- Alert band
M. Schneider — stride asymmetry +2.5 above baseline, third week inside the alert band — flagged for review
One point per training week, index points on the vertical axis — synthetic values.
Window 04
Squad readiness
22 tiles drift with synthetic readiness values while micro sparklines record each trend — once per loop, one player escalates to a flagged state. The roadmap view behind it: Match Readiness.
Starting XI
- GK L. Fischer91%
- RB J. Walker88%
- CB T. Moreau90%
- CB Á. Ortega86%
- LB H. Clarke83%
- DM N. Lefèvre89%
- CM R. Vázquez85%
- CM M. Schneider66%
- RW D. Bennett87%
- ST A. Girard92%
- LW S. Romero88%
Rotation
- GK P. Hartmann90%
- DF O. Dubois85%
- DF C. Navarro82%
- DF E. Whitmore88%
- MF F. Keller86%
- MF G. Marchand81%
- MF I. Serrano89%
- FW T. Brandt84%
- FW W. Ellison87%
- FW J. Fontaine83%
- FW L. Cabrera90%
M. Schneider — readiness −18 in 48 h · stride asymmetry drifting — flagged for review
Flag thresholds borrow their baseline-vs-drift logic from Injury Risk Signals.
Window 05
Running gait
A side-view runner while cadence, ground contact, and asymmetry settle to their values — the same skeleton reading, pointed at one athlete instead of a squad. The product surface behind it: Personal Performance.
Cadence
176 steps/min
Stride length
2.42 m
Ground contact
L 218R 224 ms
Flight time
120 ms
Contact asymmetry
2.7%
Vertical oscillation
7.1 cm
Footstrike timeline · last 3 s
Synthetic runner — the readouts settle as the stride stabilizes, then hold.
Phone-camera skeletons arrive noisy and occluded; Skeleton Repair & 2D→3D Lifting cleans and lifts them before readouts like these are computed.
Window 06
Return to play
One rehab athlete’s stride asymmetry descends week by week into his own pre-injury baseline band, and the jog, change-of-direction, and sprint milestones clear as the week cursor passes them. The research program behind it: Return-to-Play Monitoring.
Stride asymmetry — rehab sessions vs. pre-injury baseline band
- Rehab sessions
- Pre-injury baseline band
- Milestone
One point per rehab week — synthetic values; milestones clear as the cursor passes.
Rehab week
12/12
Current asymmetry
2.6%
Personal baseline band
1.8–3.0%
- W04Jogging cleared
- W07Change of direction cleared
- W10Sprint cleared
Window 07
Predictive runs
A dashed forecast extends 2.5 seconds ahead of one moving player, and the solid actual path catches up to it — probabilities, not certainties. The research direction behind it: Predictive Runs.
- Forecast
- Actual path
Window 08
Movement identity
Two clips from different sources run with the same movement signature; the match meter fills and a three-name shortlist locks its top candidate — the system shortlists, people confirm. The research direction behind it: Identity from Movement.
Footage A — broadcast · far
Footage B — training camera
Signature match0.93
Shortlist
- E. LaurentLOCKED0.93
- K. Bauer0.71
- D. Álvarez0.58
A shortlist, not an identification — people confirm every match.
Window 09
Ask the knowledgebase
A coach asks a real analysis question; the knowledgebase answers with numbers, a caveat, and an evidence line — two exchanges typing themselves out. The long-form scenarios live on the AI-agents page.
Saturday’s opponent presses high for the first 20 minutes, then drops into a mid-block. Which of our build-up patterns held up against that kind of press this season, at what completion rate — and where exactly do we lose control if their right side overloads?
Three of your five recurring build-up patterns survived a sustained high press. The third-man release through N. Lefèvre kept an 81% completion rate across 41 pressed sequences (team median under press: 64%), and the goalkeeper’s diagonal to J. Walker held in 9 of 11 attempts (82%). Under a right-side overload, control breaks at H. Clarke: his forward option disappears within 2 seconds in 7 of 9 overload windows. One caveat — only 3 of the 12 analyzed matches faced a press this sustained, so treat the ranking as directional.
evidence · 12 matches · 41 pressed build-up sequences · patterns: third-man release, goalkeeper diagonal
And their pressing triggers — what sets the 20-minute spell off, how often, and does it survive their first substitution window?
Two triggers dominate: a back-pass to your goalkeeper starts 14 of their 23 recorded press waves, and a slow switch through your deepest midfielder starts 6 more; the remaining 3 have no stable trigger. The spell fades after their first double substitution: 17 of the 23 waves come before it, only 6 after — and their recovery sprints lose 0.4 m/s. Post-70th-minute data covers only 4 of the 6 matches, so read the fade as a tendency, not a rule. All 23 trigger clips are queued in your review list.
evidence · 6 matches · 23 press waves · pattern: back-pass trigger
What you're seeing
Synthetic data, real way of seeing.
Every pixel in all nine windows is generated in your browser: the query clip and its matches are drawn variations of one synthetic move with invented similarity scores, the pressing cluster follows a script, every curve on the risk and rehab charts is hand-shaped, every readiness value is scripted drift, every player name is invented, the runners are procedural, the forecast and its agreement value are staged, the scouting shortlist compares two clips we generated, and both knowledgebase exchanges were written by us. No real footage, tracking data, or model output is streamed here.
Go deeper
From demo to platform.
What is real is the way of seeing: the production pipeline reconstructs skeletal poses from actual footage and feeds both product surfaces — the club console these windows sketch and the personal app the gait check points to.