Live view
Picture delivery in this browser
Measured here, from the pictures this page has shown; never sent anywhere.
Recognition
Incidents
No loaded incident matches these filters. Only the latest incidents are loaded on this page; older ones are not searched.
Ask about the live picture
Answers come from the on-device scene model and describe one frame: they are advisory. Chat cannot acknowledge incidents, change rules, control the camera, run commands or identify anyone.
Questions about what was recorded (“what happened in the last 10 minutes?”, “who is in the room?”, “were any alerts triggered?”) are answered from the system's own records, not by the scene model, and say what they cover.
Experimental. A zoomed answer is about a crop cut from the held picture's own recorded pixels (the main stream's where there is one), not an enlargement of the small picture. It can show no more than was recorded: what is unreadable in the recording stays unreadable. A scan asks about four overlapping parts of one held picture in turn: it covers what the camera saw in that picture, not areas out of view, and not a camera movement. A held picture is kept for a short time only and is never saved.
Features
These controls change what the runtime does, for every viewer. Appearance changes only this browser. Core monitoring (the camera, detection, tracking, rules, incidents and alerts) has no switch here, and nothing is loaded, admitted or restarted from this page.
Appearance
Saved in this browser (with JavaScript); without it, the choice is kept in this page's address. States keep the same meaning, label and icon in every palette; camera pictures and snapshots are never recoloured.
The live picture's size and the optional panels are this browser's only: they change no feature (a hidden panel's feature keeps running; use Features for that), and the freshness bar, NOT CURRENT and any warning a hidden panel holds always stay in view.