03 SPOTTER

The right flyout,
already cued.

The booth gets onto a player who just scored. Somewhere in your operator's playout list is the perfect game-line flyout. Spotter hears it, resolves it against the game state, and arms that flyout in their own engine — reason included. They take it with the key they've always used, or it quietly expires.

Talk to us about your graphics stack →

YOUR ENGINE IS THE ONLY SURFACE

No new screen. No new panel.

Spotter's defining move: suggestions appear as a cued item inside the engine the operator already runs — a Take Item in XPression's Sequencer, a queued element in PRIME, a cued page in Viz — populated, labeled, with a plain-language reason riding in the description field. It never reorders what the operator has already cued. Take it or ignore it; ignored suggestions expire and clear themselves.

Your graphics engine — playout queueON AIR
CUED Opening billboard
CUED PP stats panel
ARMED Player game line — #93 just scored (2nd) · 2 pts
CUED Shots on goal
CUED Out-of-town scores

HOW IT WORKS

Listen to the booth

An edge unit on the production LAN takes a clean mix-minus of the announce audio and runs true streaming speech-to-text — finalized words in well under a second.

Resolve the moment

A trigger engine seeded with the game's rosters — names, aliases, numbers — fuses each mention with live game state. A name right after a goal is a strong suggestion; the same name in dead air may not clear the confidence gate at all. Game state can originate suggestions with no audio — a goal arms the scorer's line before the call finishes.

Arm in your engine

The chosen flyout — one of your existing, already-licensed templates — appears as the operator's next take, populated and labeled with the reason. One suggestion at a time, never a list.

The operator takes

Same key, same muscle memory. Spotter arms; the operator takes; nothing ever fires itself. Every failure mode degrades to "no suggestions" — never to a blocked or misfiring scorebug.

A STEERING LAYER, NOT A DATA LAYER

Nothing about your graphics or data changes.

Spotter never supplies, renders, or owns a statistic. It only chooses which of your existing flyouts to arm — and those fill themselves from the data feed you already pay for. No new data licensing, no stat-accuracy liability, no change to what your broadcast looks like. Only how fast the right graphic gets cued.

// operating philosophy
01  one suggestion at a time — never a list
02  every suggestion carries a legible reason
03  dismissing is as cheap as taking
04  below confidence = silence. silence earns trust.
05  local on the LAN — no cloud dependency at air
06  zero air incidents, zero queue disruption. ever.

BUILT FOR THE ENGINES IN THE ROOM

Setup is a wizard, not a project.

Rack the edge unit, connect three things — your graphics engine (Spotter discovers your actual flyout templates by name), the roster, and the announce audio — map trigger types to templates, run the readiness check, and it's ready for air. An optional feather-light operator strip runs on stock truck Windows: no admin rights, no runtime install.

Ross XPression

Suggestions arm as Take Items in the Sequencer — the operator's next take, in the queue they're already running.

Chyron PRIME

Queued elements, populated and labeled, inserted without touching what's already cued.

Vizrt

Cued pages with the reason in the description — same behavior, third engine, by design.

QUESTIONS CREWS ASK

Is this replacing my operator?
No — it can't, structurally. Spotter only arms suggestions; a human takes everything to air. The point is compressing the hear-it, remember-the-template, find-it, populate-it, cue-it loop that exhausts even great operators — so the flyout that used to be eight seconds late airs on time.
Will it touch my cue or reorder my takes?
No. Spotter inserts one suggestion as your next take and never moves, reorders, or clears anything you've already cued. If it can't place a suggestion cleanly, it doesn't place one — your stack stays yours. Your own take key is the only thing that puts a suggestion to air.
Can it actually hear names over the crowd?
Spotter listens to a clean mix-minus of the booth — the announcers' feed, not the crowd mic — and the recognizer is seeded with the game's rosters, so it biases toward the names, numbers, and aliases that matter that night instead of open-vocabulary guessing.
What happens if it fails mid-game?
Suggestions stop appearing. That's the entire failure mode — your scorebug, your queue, and your show are untouched, because Spotter was never in the path between your operator and air. Game-night protocol is on the support page.
Does it need internet during the game?
No. Everything runs locally on the edge unit on your production LAN. There is no cloud dependency at air time.
Where does the stat data come from?
From the feed you already license. Spotter arms your existing flyout templates and they fill themselves exactly as they do today — Spotter never supplies or modifies a statistic.
Our truck PCs are old and locked down.
Expected — that's why the ML runs on the shipped edge unit, and the only thing that optionally touches a truck PC is a lightweight strip built for stock, GPU-less, no-admin Windows.
What do I need enabled on my graphics engine?
Spotter drives your engine through its standard control interface — the DashBoard API for XPression, MSE for Viz, the control API for PRIME. We confirm exactly what's licensed and enabled on your rig during setup, before anything goes near air. The full picture is on the architecture page.

Hear the booth think.

Live demo with recorded game audio — watch the right flyout arm itself in the queue before your ears catch the cue.

Talk to us about your stack →