Incident Reconstruction: the timeline, assembled

What it does

Svid has already described every frame on every camera. Incident Reconstruction reads that whole timeline and pulls the moments relevant to one incident into time order. Ask about it in plain English, "the collision in the yard around 3pm" or "the argument at reception on Tuesday", and get the sequence back: the lead-up, the event, the aftermath, each entry linked to its exact moment in the footage.

Across cameras, not just within one

Incidents move. Someone enters on camera one, crosses the floor on camera three, and the event happens on camera six. Every camera's frames are described and time-stamped, so searching each folder for the same description, "the person in the high-vis vest, between 2:40 and 3:10", pulls the relevant moments from all of them into one time-ordered account. Svid deliberately does no people tracking or re-identification: the cameras don't follow anyone. A person describes what they're looking for, and the search finds the matching moments on each camera separately.

What you get

A time-ordered set of described moments with links into the footage. It reads like the account you'd have written after an afternoon of scrubbing through four windows side by side, produced in minutes instead. Notable findings can go out by email, and the JSON webhook carries events into a claims or case system if one exists.

Where it runs

Incident Reconstruction is switched on per camera and draws on whichever ones it's enabled for. It works on RTSP feeds, uploaded files and video URLs. Recorded footage is the normal case here: the incident already happened, the export goes in, and the same pipeline runs as it would on a live feed.

What it costs

Report checks are the cheap end of the Store. Before you buy, Incident Reconstruction shows a live price computed from your own usage as "+$X/mo at your usage", and report checks start from a $3 a month floor. It joins a subscription priced from your own usage, with a $9 a month minimum, and the whole derivation is on screen at signup. The pricing section shows the arithmetic.

How long does a reconstruction take?

Asking takes about as long as typing the question; results come back in seconds. The afternoon of side-by-side scrubbing is the thing this replaces.

Can it follow one person across several cameras?

No, deliberately. Svid does no people tracking or re-identification. What it does instead is search every camera for the same description, "person in the high-vis vest", and the matching moments from each come back in time order, which is what a reconstruction actually needs.

Can I run it on footage from before we had Svid?

Yes. Upload the export; recorded footage goes through the same pipeline as live, so an old incident can be reconstructed once the footage is in the account.