About the cam

The camera is a pan-tilt-zoom unit fixed on the open water. An AI tracker watches for boats and follows them on its own, so there's almost always something happening on screen.

The tracker's eye is a computer-vision model that started life as an off-the-shelf YOLO detector and was then custom-trained for this exact view: thousands of images of boats on Lake Travis, captured by this very camera, each one human-verified as actually being a boat — and not a dock, a boat parked in a dock, a bird, a tree, or a building across the cove. That extra training is what lets it pick out small watercraft at long distance across open water.

When you want to look somewhere specific, you take control: tap a named spot — the dock, Captain Pete's, Graveyard Point — or type the same commands in the YouTube live chat. After a short idle, the camera hands control back to the tracker.

The view stays on the lake. The camera is aimed at — and physically limited to — open water and named public landmarks; its range can't reach private docks, yards, or windows.

The AI tracks boats, not people: no faces are identified or stored. If a restart takes the stream offline for a few minutes, it comes back on its own.

The lake level, weather, and alerts under the cam come from public authorities — LCRA and the Texas Water Development Board for the reservoir, the National Weather Service for forecasts and watches, and AccuWeather for current conditions — refreshed every few minutes.

We show a value only when we can source it honestly; anything we can't, we leave out. Every figure is cited in full on the lake-level page.

The cam runs on a professional EmpireTech PTZ camera and Frigate, an open-source NVR doing the on-site AI detection.

The cloud side is all AWS: Amplify Hosting serves this site, Lambda assembles the lake and weather data every few minutes, and the AI captions in the live feed are written by Amazon Nova on Amazon Bedrock — one small model reading the camera so you don't have to.