The platform

Everything your cameras saw, in words.

OmniVision turns video into a written record you can search, question and act on. Here is what that looks like in practice, from the first search to evidence that holds up.

Get started Watch the demo

Describe it. Find it.

Every scene gets a written caption, a set of detected objects and a search embedding. Search combines all three, so "a man in a red jacket carrying a box" finds the moment, not just every man.

Times are real filters: "after 9pm", "last Tuesday", "this morning"
Results ranked by relevance, with the clip, the camera and the time
Search one camera, a group, or every source including archives

New · Ask OmniVision

A conversation with your footage

Ask a question and get a short written answer, built only from what your cameras recorded. Each statement carries a citation you can click to watch the moment it came from.

Follow-up questions keep the context: "and after that?"
Upload a video and ask about it straight away
Counting questions are answered from the vehicle counter, not guessed
Runs on a local language model. Your questions never leave the site

Ask OmniVision · Warehouse

Summarise what happened at the loading bay between 6 and 9 this morning.

The bay was empty until 06:40, when a forklift moved two pallets to the door S1. A white truck reversed into bay 2 at 07:30 S2 and two people unloaded it until about 08:10 S3.

Was the door left open afterwards?

Yes. The roller door is still open in the last clip at 08:52, with no one nearby S4.

Example conversation for illustration.

Alerts & incidents

Rules you can read out loud

Write what you want to know about as a sentence. OmniVision shows how it understood the rule (which object, which camera, which hours) before it ever fires, so there are no surprises at 3am.

Push notifications to phones and browsers
Incidents open on their own and collect the related clips
A daily digest of the last 24 hours, written for people who weren't watching

Vehicle counting

Draw a line. Count what crosses.

Pick a frame from a camera or an uploaded video, draw a counting line, and OmniVision tracks every car, truck, bus and motorcycle across it, once each, with its direction.

Totals by type and direction, with a thumbnail for every crossing
Ask "how many cars passed in the last two hours?" in chat or search
Works on live sources and on recorded or uploaded video

Gate line · example

39Cars

6Trucks

2Motorcycles

Dashcam telemetry

Where, when and how fast

OmniVision reads the GPS track embedded in dashcam footage. Every trip is drawn on a street map, with a speed profile and driving events pinned where they happened.

Hard braking, sharp cornering and speeding events
Click a place on the map to find every clip recorded near it
Maps are stored on the box: viewing a trip never contacts a map service, and works offline
No extra tracker in the vehicle

Reports & evidence

From question to signed report

Turn search results into an incident report with the clips attached. Exports are signed and come with a hash manifest, so anyone can check later that nothing was changed.