Live-Stream Hype Engine
A real-time monitor that watches a Whatnot live stream, reads sales off the screen with OCR, and fires synchronized OBS overlays, smart-light effects, and sound the instant a card sells.
The Challenge
Selling cards on live streams is part commerce, part show. Every sale and every "win" deserves a moment—but a solo seller cannot watch the stream, run the camera, and trigger celebrations by hand all at once.
The Approach
I built a Python monitor that screen-captures the live stream at ~15fps and watches specific zones: a brightness drop signals a win, and OCR on the sale bar reads the card title. It matches that title against the current F1 driver grid, then fires three systems in one coordinated burst—a driver-specific OBS overlay over WebSocket, a Govee smart-light color effect over REST, and a celebration sound—all guarded by cooldowns and error isolation so streaming chaos never cascades into a crash.
The Outcome
Sales and wins trigger instant, driver-aware hype automatically, turning a one-person stream into a produced show. A pure presentation layer, cleanly separated from the commerce backend.
Key Details
- Multi-zone screen capture at ~15fps with auto-calibration
- Win detection via baseline-relative brightness drop
- OCR (macOS Vision framework) reads sale titles, fuzzy-matched to the F1 grid
- Coordinated firing: OBS overlay + Govee lights + sound, with per-system error isolation
- Cooldown guards prevent re-trigger spam during rapid-fire sales