Variable Sense turns any video, audio or document into answers — in the moment it arrives, or years after it was captured. Video comes first, audio and documents are planned.
Briefings open soonRuns entirely on your own infrastructure — on-premise or in your cloud account.
Answers sit scattered across footage, calls, photos, files and sales records, and no one can read them all. We start with video: any source, live or recorded, searchable in a sentence.
The first product: video analytics
Live streams and recorded files, from any source. The same detection, search and alerts work on footage from a fixed camera, a dashcam, a body camera, a drone or a phone. People only look when it matters.
Zone and behaviour rules need a camera that stays still. Everything else works on moving footage too.
The layer everything else runs on: finds and follows every vehicle, person and object, in any video — recorded files or live streams.
Reads number plates, traces each vehicle's route across your cameras by its plate, and monitors traffic violations.
Clothing, accessories, apparent age and gender, plus opt-in face recognition that follows enrolled people across cameras.
Someone entering a restricted zone, staying too long, running or gathering in a crowd, flagged on any fixed camera, with heatmaps of where people and vehicles go.
Filter by camera, time, colour or plate, describe it in a sentence, or replay the route a vehicle or enrolled person took.
Alerts from any analysis you switch on, signed and delivered to your systems, with a live dashboard and an API for your own tools.
Two jobs, on the cameras you already own.
Live
Models run continuously across every feed and flag only what needs a person: wrong-way driving and vehicles in a lane they are not allowed in, congestion building up, someone entering a restricted zone or staying too long, and people you have enrolled arriving at a camera. Traffic, congestion and face alerts reach your systems in seconds, with a snapshot attached.
Recorded
Every object is recorded as it is seen, so investigation becomes a query. Find a white truck on the ring road between 2 and 4am, a person carrying a box in the last two hours, or every camera one number plate passed. Results in seconds, with the best image of each.
Attributes
Search works on attributes: vehicle type, colour, plate, clothing, direction, time and camera. Face recognition is opt-in, switched on per camera, and only finds people who have been enrolled with their consent.
The platform
Variable Sense is built to read everything an organisation records. Video is where the gap is widest, so video is where we started. Audio, photos, documents and business records come next, and every source resolves to the same entities — person, vehicle, object, event — so one question can reach across all of them, through the same interface, on the same deployment.
Planned, and none of it is built yet. Each part will run on the same deployment as video and feed the same people, vehicles, objects and events.
Planned
Separates each speaker in a call or recording, writes down what was said, and checks whether two recordings are the same voice. It also picks out the sounds behind the voices — traffic, horns, sirens, trains, crowds, glass breaking, gunshots — which can place a recording in time and space, or contradict where a caller says they were.
Planned
One identity built from every clue: a face, build and clothing, a voice, and the names and phone numbers in the records. Each match comes with a score and the evidence side by side, for a person to confirm. Today, faces are matched across cameras.
Planned
Photos, witness statements, FIRs, chat exports and scanned papers, read for names, places, dates, vehicles and phone numbers, and linked to the footage those appear in. A whole case file becomes one thing you can question.
Planned
Sales, reviews, bookings, transactions and logs, read alongside the footage. A restaurant could ask why Friday dinners dropped in March and see the answer in the bills, the reviews and the kitchen camera together.
Planned
Every file in a case, questioned together. Today: CCTV, dashcam, body-camera and phone footage, searched by person, vehicle, plate or a description, with journeys traced across cameras. Planned calls, photos and case files in the same case, and questions across all of it.
Approaches, crossing points and fence lines, watched around the clock. Anyone entering a restricted zone, lingering near the fence or running is flagged on fixed cameras; every person and vehicle is followed, vehicles are identified by plate across cameras, and enrolled people are matched by face. Drone footage is read the same way.
Substations, ports, water treatment, plants and logistics hubs. Continuous awareness across large sites without adding headcount, with restricted areas and fence lines watched by zone rules. Works with the systems already installed — today, the cameras.
Traffic authorities running city and highway networks. Wrong-way and wrong-lane driving, congestion as it builds, and every vehicle traced across cameras by its plate. Dashcam and fleet footage can be searched the same way.
Malls, shops, restaurants and hotels. How many people come through, where they go and how long they wait, measured on the cameras already up. Planned the same questions answered with sales, bookings and reviews alongside.
Dashcam and phone footage from a claim or a trip, searched for vehicles, plates, people and moments. Planned claim forms, repair bills and GPS logs read alongside, to check that the footage and the paperwork agree.
Engagements will begin as a paid pilot on a single site, measured against one number: response time, or hours saved per investigation.
Coming soon
We are getting ready for our first pilot sites. Until requests open, you can write to us at contact@