Ask Your Footage a Question
Three recordings of half an hour each is ninety minutes nobody is going to watch. Describe the moment you are looking for and Trinocular AI returns the camera, the time and the clip.
Recorded Is Not the Same as Found
Every site already keeps more footage than anyone can watch. The incident is usually in there.
A camera that records for a shift produces hours of video, and a site with sixteen of them produces more in a day than a team will review in a month.
So the footage gets kept and not watched. When something comes up, a stolen tool, a damaged vehicle, a complaint about who was where, someone is sent to scrub through a recording with a rough idea of the hour. Most of the time they find nothing, not because it is not there, but because looking for a few seconds inside a few hours is a bad use of a person.
Search changes what the archive is for. Instead of a thing you keep in case you ever need it, it becomes a thing you can interrogate in the time it takes to type the question.
Index Once, Ask Anything
The work happens on the way in, so the question is answered when it is asked.
- 01
Point it at the footage
Upload clips, or connect the recorder you already have. Trinocular AI reads every frame once and builds an index of what is in it: people, vehicles, plates, PPE, fire and smoke, and which zone each was in.
- 02
Ask in your own words
Type the moment you are looking for the way you would describe it to a colleague. Not a filter, not a query language, and not a list of camera numbers and times you would have to know in advance.
- 03
Get the frame, not the file
Results come back as moments: the camera, the timestamp, the zone and the few seconds either side. You open the clip at the point that answers the question instead of scrubbing a recording to find it.
Written the Way You Would Say It
No query language, no camera numbers you would have to know in advance.
-
Show me anyone who entered the yard without a hard hat yesterday
PPE compliance, filtered to one zone and one day
-
Which vehicles were turned away at the gate this week?
Plate recognition, crossed with the access decisions
-
Was there any smoke near the fuel store overnight?
Fire and smoke, over the hours nobody was watching
-
Find the white pickup that came in around four
A vehicle by description, not by plate
-
Did anyone go near the crane while it was lifting?
Zone entry, during a window that matters
-
Who was last at the rear door before it was forced?
Working backwards from a thing you found later
Footage Search is a module in the command centre. Try it on the live demo
Questions
What is the point, when we already keep the footage?
Keeping footage and being able to find something in it are different problems. Three cameras recording a shift is twenty four hours of video to review for one question. Almost nobody does that review, which means an incident that was captured is still an incident nobody found. Search is what closes the gap between recorded and known.
Does it only work on footage you have already alerted on?
No. Alerts fire on the rules you configured; search works across everything indexed, including the hours where nothing was configured to alert. That is usually where the question you did not expect to ask gets answered.
What happens to footage we upload?
The same deployment choice as the rest of the platform. Sites whose footage cannot leave the premises index locally and nothing is uploaded. Where footage is processed in the cloud it is held under the retention you set and deleted on that schedule.
How accurate is a search?
A search is only as good as the detections underneath it. It can find what the models can see: people, vehicles, plates, PPE, fire and smoke, and zone crossings. It cannot find what the camera could not resolve, and a question about something outside the configured detections will say so rather than return a plausible wrong answer.
Can we try it on our own footage?
Yes. The usual first step is a short pilot on recordings you already hold, because the honest test is whether it finds something in your own archive that you did not know was there.
See it on your own cameras
Bring your existing CCTV. We will assess what your angles can actually support and show you the detections running on your own footage.