Not just detect motion. Actually understand what's in the frame.
Traditional motion detection tells you something moved. That's about as useful as a smoke detector that goes off whenever the temperature changes. You still have to watch the footage to find out what happened.
TetherX Video AI is different. It identifies people, vehicles, and animals in real time, tracks their movement across the scene, and classifies what it finds - all without a single frame leaving your site.
Real-World Results
These screenshots are from a live site in Canada. Night time, snow on the ground, well below zero.
A person tracked across an entire carpark at 91% confidence. The yellow line is their actual trajectory - you can see exactly where they came from and where they went.
Two people detected at distance - tiny figures in a wide, snow-covered scene - at high confidence. Most camera analytics wouldn't register them at all.
A result like this isn't luck on a good day. Reliability comes from exposure, so before any of this reached a customer site we ran the models against thousands of hours of footage from hundreds of cameras - daylight and pitch dark, glare and floodlight, rain, snow and fog, wide overviews and tight doorways, busy scenes and empty ones. Every awkward angle and lighting condition we could find went in, because the test that matters isn't whether it works in a demo - it's whether it works on your site, in your light, at your camera angle. That's what the hours buy: detection you can trust across the conditions real installations actually run in.
Search Your Footage Like a Database
All of this metadata becomes searchable. Instead of scrubbing through hours of footage, you filter by what matters:
- Person, vehicle, or animal - narrow down by detection type
- Clothing colour - top and bottom separately
- Vehicle colour - find that specific car
- Moving or stationary - filter out parked vehicles or loitering
- Direction of travel - which way were they heading?
"Show me everyone wearing a red jacket between 6pm and midnight." Results in seconds.
For integrators managing multiple sites, this changes the conversation with your clients entirely. Instead of "we'll review the footage and get back to you," it becomes "let me pull that up right now."
Edge Processing - No Cloud Required
The AI runs entirely on the TetherBox. No cloud processing, no footage leaving the site, no additional subscription for a cloud analytics service.
A recent-generation TetherBox handles the AI workload locally, in real time, alongside normal recording. There's nothing extra to install, no separate server to maintain, and no bandwidth impact.
For sites with strict data sovereignty requirements, this is significant: the video never leaves the premises, and the intelligence stays on-site too.
They Hear, Too
Video has one limitation: it only works where the camera can see. So we gave the TetherBox a second sense - TetherX Audio AI picks up glass breaks, alarms and screams that never cross the lens.
Try TetherX Video AI
Video AI runs as an optional per-camera service on TetherX. Natural-language search ("find the red truck", "person in yellow jacket near the entrance") lands in Q3 2026 - see the cloud / edge / camera analytics post for how we think about the analytics layer overall.
If you're an integrator who wants to stand this up on a real site, get in touch at beta@tetherx.io or reach out to your TetherX account manager.
Leave a Comment