The interaction layer for live and recorded media
Make every moment
playable.
PLLAY transforms video, streams, sports, gaming and creator content into predictions, challenges and verified outcomes — powered by autonomous interaction intelligence.
Three products. One system underneath.
PLLAY Pulse
In DevelopmentFree participation. Brand-funded engagement.
Turn live and recorded content into measurable fan interaction without asking audiences to wager.
Free creator-operated predictions run on Twitch today, and fans play them at no cost. What is still being built is Pulse as a brand-funded product: sponsor demand, approved AI-generated interactions and engagement-based pricing. There is no self-serve sponsor tooling and no external brand pilot has run.
Explore PLLAY PulsePLLAY Arena
In DevelopmentReal stakes. Verified skill.
Power eligible adult skill competitions with identity, jurisdiction and responsible-play controls.
Arena is not open. External cash rails are disabled, identity verification runs in a non-production configuration, and no cash-out path is enabled for any account.
Explore PLLAY ArenaPLLAY Vision
AvailableEvidence becomes outcomes.
Read authorized video, structure visible events and verify what happened.
Validated on recorded video. Automated live-stream settlement is not switched on; when an outcome cannot be confirmed, PLLAY holds or refunds rather than guessing.
Explore PLLAY VisionTwo participation models. The same evidence.
Pulse and Arena differ in exactly one respect: whether a stake is attached. How a moment is found, how an outcome is read, and how it is recorded are the same system in both.
Reads the evidence
Vision AI
AvailableRuns optical character recognition and reconciliation over authorized video, produces a candidate outcome, and records the confidence behind it.
Benchmarked on recorded video only. Vision AI has never run on a live stream and does not settle live pools — the creator or operator still confirms.
See the detailFinds what is worth asking
Moment Agent
In DevelopmentWatches for a moment an audience would want to call, and drafts the interaction around it.
The permission layer this needs is built and switched off. What is missing is the part that proposes — no agent has ever drafted a question on PLLAY.
See the detailRemembers what happened
Outcome Graph
AvailableEvery verified outcome, and every call made against it, recorded as a durable record of who read a moment correctly.
It records outcomes and accuracy. It is not a partner-facing data product — there is no export path or external reporting surface today.
See the detailPLLAY Pulse
No stakeA fan calls a moment for free. A sponsor funds the reward. The outcome is verified the same way.
PLLAY PulsePLLAY Arena
Eligible stakeA verified adult in a permitted jurisdiction commits credits. Identity, geography and responsible-play controls apply. The outcome is verified the same way.
PLLAY ArenaPLLAY Vision AI — Arena Dodge AUTO-settle walkthrough
Beat 1 of 11: ARENA DODGE — BROADCAST FEEDRecorded footage · PLLAY's own esport
Beats 3, 4, 5, 8, and 9 above are a real Vision AI run over PLLAY-owned Arena Dodge footage: the tracked teams, round-win read, final score, and AUTO-settle verdict are its real, un-coached output. Beats 1, 2, 6, 7, 10, and 11 are illustrative product UI demonstrating the fan and creator experience, not measured data. Vision AI runs on recorded video, not live streams — the method, the full trial record and the run log are on the validation page.
PLLAY confirms outcomes when evidence is sufficient. When it is not, the event is held or refunded — never guessed.
What we are building the system to do alone.
Five steps, end to end. 2 of them run today; the rest are being built. We would rather show you the whole pipeline with the gaps marked than describe a finished one.
Detect an eligible moment
Recognize that something just happened on the broadcast that an audience would want to call.
The permission layer this needs is built and switched off. What is missing is the part that proposes — no agent has ever drafted a question on PLLAY.
Generate the interaction
Draft the question, the options and the window, shaped to the moment rather than to a template.
No agent has drafted a question on PLLAY. Every prediction ever run was written by a person.
Open and close participation
Run the window: take calls, lock it before the outcome is knowable, and hold it until evidence arrives.
Ships today, operated by a person. This is the step that is real — the autonomy around it is not.
Verify the outcome
Read the evidence off the video, settle when confidence clears the bar, and decline when it does not.
38 recorded broadcasts, 31 auto-tier, 0 wrong auto-settles. Every trial is recorded video; live-stream settlement is not switched on.
Optimize what comes next
Learn which moments produced real participation, and let that shape what gets surfaced afterwards.
The outcome record exists. Nothing reads it back to influence what is proposed, because nothing proposes yet.
The autonomy ladder, mode by mode, with what each one is still waiting on: PLLAY Autopilot.
Pick the path that matches you.
Creators
Give your audience something to do mid-broadcast, and keep a share of what it produces.
PLLAY Pulse
OpenBrands and agencies
PLLAY Pulse lets a brand fund something an audience wants to do, inside content they already chose to watch. No purchase, no entry fee, and no ad break.
PLLAY Pulse
OpenMedia and platforms
Make live and recorded content something your audience plays, inside your own product and under your own brand.
PLLAY Vision
OpenGame publishers
PLLAY Vision reads outcomes off authorized video — per-title, with an adapter model and a published accuracy record you can inspect before you believe it.
PLLAY Vision
OpenReal-time content and recorded libraries.
The same pipeline reads both. What differs is how much of it runs without a person, and that is stated per column rather than averaged into one claim.
Live
A creator opens participation mid-broadcast and the window closes before the outcome is knowable. The creator or an approved operator confirms what happened.
The creator or an approved operator confirms the outcome. That confirmation is what releases a pool.
Benchmarked on recorded video only. Vision AI has never run on a live stream and does not settle live pools — the creator or operator still confirms.
Replay
A recorded library is addressable the same way. This is also where the verification record comes from — every published benchmark trial is recorded video, not live traffic.
Verified when confident. Held or refunded when unclear.
Replay participation is not a self-serve product. Running it against a library is a scoped engagement with our team, not something you can switch on.
Every trial, including the ones where the system declined to determine an outcome, is published at the validation record.
Watching tells you what people saw.
Participation tells you what they believed.
Every PLLAY interaction creates a verified signal:
- •what the fan predicted
- •when they acted
- •what content they understand
- •how accurate they are
- •which creators, competitions and formats bring them back
Watching produces impressions. Participation produces evidence — a record of what an audience actually believed and what turned out to be true. Most of that record carries no money at all: a free call on PLLAY Pulse produces the same verified outcome, and the same signal about who read the moment correctly, as a staked one on PLLAY Arena. The Outcome Graph is not a product PLLAY sells today. It is the layer underneath the products, and it gets more valuable as verified outcomes accumulate — whether or not a stake was ever attached to them.
The loop is being built to close. Each turn makes the next one cheaper — but it turns on participation volume, and that volume is still early.
- More creators and partners
- More playable moments
- More fan participation
- More verified outcome data
- Better personalization and intelligence
- Higher engagement and monetization
- back to more creators and partners
PLLAY is in a demo phase. Contests, settlement and every control described here run end to end against credits, but external cash rails are switched off at the database level — no real money enters or leaves the platform today. What we do and don’t claim.
Built to plug into what you already run.
The same prediction engine, identity and settlement stack shown above is offered to partners three ways — each at its own real stage, not one blended claim.
Stage 1 · Available
API
Prediction API, partner-key auth.
Real and documented — a linked partner launches and settles predictions today. Offered to partners today through pilot access, not a public self-serve API.
Stage 2 · Available
Embed
A chrome-free embed, themed to your brand.
Renders from your brand configuration and runs inside your own product today. This is the half of the old white-label card that actually ships.
Stage 3 · In Development
White-label
Full white-label integration is roadmap, not yet piloted.
Production has zero partner deployments and zero partner-creator links — there is no pilot to describe. The deeper integration most partners actually want is still hands-on.
Stage 4 · In Development
SDK
No published client library in any language yet.
The API above is callable directly; a packaged SDK wrapping it hasn't shipped.
Start here · Available
Vision Proof Sprint
A scoped engagement, run end-to-end with our team.
Available to book today. No completed external sprint has published results yet, so this is an offer with a defined scope — not a case study.
Your audience is already watching.Give them something to do.
Free predictions run on Twitch today and Vision is verifying outcomes today. Tell us what you want to make playable and we will scope the smallest thing that proves it.