The camera becomes the official.
It watches. It counts. It verifies. PLLAY AI Referee transforms observable creator performance into structured, verifiable outcomes.
One system, two names — this is the stadium view of Vision AI, not a second one. Every claim on this page traces back to the bench record published there.
A challenge, read live.
The walkthrough below runs a pushup challenge because it is the clearest illustration of the shape. It is a concept demo: nothing verifies pushups today, and the demo says so in every frame.
- CHALLENGE LIVE
- AI REFEREE LIVE
- RESULT VERIFIED
Prediction state — SETTLED
PUSHUP CHALLENGE
Result verified
41 valid · 3 invalid
Target 40 — verified ✓
Credits, not dollars. PLLAY is pre-launch and cash rails are off.
Verification breaks into seven primitives.
This is an architectural description of what verification requires in general — not seven shipped models. 4 of the 7 run today, and only for gaming; the rest are exactly what a pushup, a free throw, or a sprint would need, and none of them exist yet.
Track human movement.
No pose model exists in the adapter catalog. Nothing watches a body today.
Identify relevant balls, targets and equipment.
A presence probe over the regions a gaming adapter declares — shipped, benchmarked, published at Vision AI.
Count valid repetitions and actions.
No rep- or action-counting model exists in the adapter catalog.
Follow movement and trajectories.
No ball- or object-tracking model exists in the adapter catalog.
Measure event duration.
A clock region, read alongside the score on scorebug-and-banner adapters — shipped for gaming.
Convert observations into structured results.
The reconciler's job: frame reads become one candidate — score, winner, confidence tier.
Return the result to PLLAY's rules and settlement engine.
The settlement gate exists and checks the tier. It proposes; a gate disposes — see Vision AI's roadmap.
The four that ship are published in full, per game and per tier, at Vision AI.
One referee. Thousands of possible skills.
The same primitives could in principle read any objectively measurable performance. Today they read one of the categories below — every other row names the specific model that does not exist yet, because a roadmap entry without a stated ceiling is a promise.
- GamingLive
The one shipped adapter catalog. Benchmarked on recorded video, published in full at Vision AI.
- PushupsNext
No rep-counting model. Nothing in the adapter catalog watches a body.
- SquatsNext
Same missing model as pushups — rep validity from a camera has no adapter today.
- Pull-upsNext
No rep-counting or bar-clearance model.
- Free throwsNext
No shot-detection model. The catalog reads game scoreboards, not real hoops.
- Soccer shotsNext
No ball-tracking or goal-detection model, and no way to read a physical pitch.
- Trick shotsNext
No object- or trajectory-tracking model for an arbitrary physical trick.
- SprintsNext
No timing or motion-capture ingest. Nothing reads a stopwatch, a wearable, or a track.
- Reaction testsNext
No timing harness a creator could run on a broadcast.
- Future skill categoriesPotential
An open bucket, not a promise — nothing here has a name yet, let alone a model.
9 of 10 categories above are still just ideas. No shipped-on date until the verification is real.
Vision AI sees the moment. AI Referee calls the result.
The record behind every claim on this page — every trial, per game and per tier — is published in full, not summarized here a second time.