- Local inference
- Detection on every stream at the property: people, vehicles, animals, packages. Nothing is sent to a cloud service unless you ask for it.
- Object states
- Not just presence: a car parked versus arriving, a person at the door versus walking past, a package left versus collected.
- Zones and schedules
- Driveway, pool deck, gatehouse, service yard, perimeter: different sensitivities, different hours, different people notified.
- Face recognition
- Household and staff enrolled; an unknown face is a separate event from a resident's late arrival, and is handled separately.
- Plate recognition
- Every plate at the gate read and matched against a known-vehicle list, with no barrier tag required and no cloud lookup.
- Audio events
- Breaking glass, alarms, raised voices, a dog in distress, where they carry meaning and on the cameras that cover them.
- Search by description
- "White SUV at the gate after nine": described in words, answered from the footage across every camera at once.
- Retention by value
- Events kept by type, continuous coverage where it matters, clips exported with the metadata that makes them evidence.