Included in Core
Know that something happened — without any AI.
Motion detection turns a live camera into a record you can search by time. It runs on your own agent by comparing frames, so there is no model to download, no GPU to buy, and nothing sent anywhere to be analysed.
What you get
An event feed, on the cheapest plan
Everything below is included in Core at $12 per camera per month, and works the same on Pro and Sovereign.
An event every time the picture changes
Each detection is a timestamped event with a thumbnail, so an operator scrubbing back has something to land on instead of hours of identical footage. This is what fills the event feed on a Core deployment.
Recording that starts on movement
Clips are captured around the movement rather than continuously, so storage is spent on the moments that matter. Retention policies and litigation holds apply exactly as they do to any other event.
Sensitivity you can actually tune
One slider per camera, from "only large changes" to "a person entering frame at night". A car park at midnight and a busy shop entrance need different settings, and you set each one from what you see.
A cooldown so a busy scene stays readable
A minimum gap between events on the same camera, so a pavement at rush hour files one event rather than one per passer-by. Adjustable from one second to five minutes.
No AI, and no model to download
Detection is a frame-difference measure computed while the video is decoded. There is no neural network involved, which is why it runs on modest hardware and why a Core install never needs the AI model bundle at all.
Nothing leaves your network to detect it
The comparison happens on the agent inside your own network. What travels to the console is the event and its thumbnail, not a video stream for someone else to analyse.
How it works
Frame by frame, on your hardware
The agent decodes the stream locally
Your on-site agent pulls the camera over RTSP and decodes it on the machine you already own. Nothing is uploaded to make this work.
Consecutive frames are compared
Each frame is scored against the one before it. While the scene is static the score stays below your threshold and nothing happens — so a quiet camera costs almost no CPU and generates no events.
A change above your threshold becomes an event
When the score crosses the sensitivity you set, the agent captures the frame, records the clip, and files a motion event with its thumbnail.
The cooldown keeps the feed readable
Further movement inside the cooldown window is still recorded but does not file a second event, so one incident reads as one entry.
It keeps running whether anyone is watching or not
Motion detection is not tied to somebody having the tab open, and it reconnects by itself if a camera reboots overnight. The question it exists to answer is "what happened while nobody was looking".
Where it stops
What motion detection is not
Worth reading before you buy, because knowing the limit is what tells you whether Core is enough or whether you want Pro.
It reacts to weather and light
Rain, snow, headlights sweeping a wall and a security light switching on are all changes in the picture, so a frame-difference detector sees them. Sensitivity and the cooldown are how you tune them out — and on an outdoor camera you should expect to adjust once after watching a night of real events.
It cannot tell you what moved
A motion event says the scene changed at 03:14. It does not say whether that was a person, a fox or a flag. If you need to know which — or to search for one person across every camera — that is object detection, and it is what Pro adds.
Specification
The detail
| Method | Frame-difference (scene change) computed during decode on the on-site agent — no neural network, no inference |
| Included in | Every plan, including Core |
| Models required | None. A Core-only install never downloads the AI model bundle |
| Hardware | CPU only, on a machine you already own. No GPU, no recorder, no cloud analysis |
| Sensitivity | Per camera, 1–100, from "only large changes" to "a person at night" |
| Cooldown | Per camera, 1–300 seconds between events on the same camera |
| What an event carries | Timestamp, camera, JPEG thumbnail, and the recorded clip |
| Availability | Runs continuously, independent of whether an operator is viewing |
| Resilience | Reconnects automatically with backoff if a camera or the network drops |
| Retention | Standard retention policies, litigation holds and audit apply |
| Upgrade path | On Pro, object detection replaces motion on that camera — the same feed, classified as a person or a vehicle |
Start on Core. Move to Pro when you need to know what moved.
15 days free on up to 5 cameras, no credit card.