Available
Fire and smokefrom EUR 2,900
89 clips, 10,822 annotated frames. Smoke and fire, plus steam and fog as false alarm look-alikes.AI video analytics for your CCTV Your cameras saw everything.Now you can ask them.
Xentim does not just see objects. It understands what is happening in the frame: it checks every alarm before a person sees it, describes the event in a plain sentence and drafts the report. You ask your footage questions the way you would ask a colleague. All of it locally, on the cameras you already have.
Built for those who have cameras but no time to watch them
property managerssecurity companieswarehousesshopsconstruction sitesmonitoring centres
Every object on its own. Live.
Live, street camera
Boxes around people and cars are only the start. A second model checks every alarm and says in words what is really there.
See how it worksOur own footage, processed by Xentim. Faces and number plates blurred automatically.
Why Xentim
Detection is standard now.
Understanding is not.
Cameras have drawn boxes around people and cars for a decade. New vision language models can do more: understand the scene, judge an alarm and describe an event in words. Xentim does this locally, on your cameras.
Built-in camera analyticsXentim
Detects motion and line crossingUnderstands what is in the frame and says it in a sentence
Finds only what was set up in advanceAsk about anything, even without a rule
Every movement in a zone is an alarm for the operatorA second model checks the alarm before a person sees it
The operator writes the report from scratchThe incident report is waiting for approval
One camera, one brandAll cameras at once, any brand, no cloud
The same model on every siteA model tuned on footage from your site
The Xentim platform
Sees. Understands. Answers.
A fast detector tracks every object, and a vision language model judges what matters. The result is a system that filters out false alarms, answers questions and writes reports.
01 / Second opinion on alarms
Before an alarm reaches a person, someone has already looked at it.
Movement in a zone triggers an alarm. Within the same second a local model looks at the frame, checks whether the box really holds what was reported and adds a one sentence description. Alarms judged false can be held back entirely. The check takes under a second, and the alarm goes to e-mail, Telegram or a monitoring centre, just like from an alarm panel (SIA DC-09).
Confirmed
Person in zone, courtyard, 11:09
A person in a green hoodie walks along the garages.checked in 0.6 s, locally
Likely false
Movement in zone, car park, 03:42
The box holds a cat, not a person. The alarm was not sent.checked in 0.6 s, locally
Uncertain
Vehicle in zone, gate, 22:15
The box covers headlights in the rain. Hard to judge, so the alarm goes to the operator.checked in 0.6 s, locally
Example verdicts. The first comes from our demo installation, the other two are illustrative.
02 / Search
Type what you want to see. Get that moment.
You do not set a time or a camera number. You describe the event in a plain sentence, and the system shows the matching clips, best match first. Click an example or wait for it to change by itself.
03 / Incident report
The report is ready before you could write it.
Any alarm or any passage becomes a draft report: the sequence with times, frames from the footage with faces blurred, the object's track and a written description. The operator edits the description, approves it and sends it to the client, the insurer or the police as a PDF or an e-mail.
Incident reportdraft for approval · courtyard camera
Sequence
- A person enters the frame from the gate side
- Zone alarm by the garages, AI verdict: confirmed
- The person leaves the frame
Description
A person in a green hoodie spends about 7 minutes by the garages in the courtyard, then leaves the frame.
Frames from the footage
PDF or e-mail once approved
Illustrative example using frames from our demo installation.
04 / Foundation: live detection
Many objects at once. Each with its own track.
The system tells apart people, passenger cars, vans, trucks, buses, trams, bicycles and wheelchairs. Every moving object gets a box, a type, a colour and a track line, while parked cars are only outlined. The faces and number plates in these recordings were blurred by the system itself.
personpassenger carvantruckbustrambicyclescooterwheelchairdog
Datasets
Events you will not find in any catalogue.
Detectors learn from millions of photos of people and cars, but they almost never see smoke at night, a fall on a platform or a fight at an entrance. We produce that footage ourselves: synthetic scenes with every frame fully labelled, with no real people and no personal data. Teams that train their own models buy them.
Available
Fightsfrom EUR 3,000
1,045 events, 12 cameras per event. 10 kinds of fight and 10 look-alikes such as hugs.Available
Fallsfrom EUR 18,000
1,009 events, 12 cameras per event. 5 fall types, fall phases and impact time.Available
Railway platformfrom EUR 3,000
1,142 events, 3 station scenes. Falls and pushes onto the track, people on the track.Coming soon
Care room
Getting out of bed and falls by the bed, ceiling camera. For elderly care and hospitals.Entirely computer-generated data. Perpetual, non-exclusive licence for recognition models. Bundle of the three skeleton datasets: 25% off.
Use cases
Where Xentim
saves the most time.
Six sectors, six concrete situations. Each piece shows what happens on site and what video analysis changes.
Plants and warehouses
Night shift, two people, forty cameras
An alert when someone is at the fence after hours, and shorter truck waits at the docks.
Read
Housing and property
A sofa bed by the bins and four days of footage
Who dumped the bulky waste, whether the bin lorry came, who blocked the fire lane.
Read
Municipal waste
The trail camera took the photo. Who will look at it?
Illegal dumping and disputed collections. The culprit without scrubbing through footage.
Read
Railways
A car stalls on the tracks. The signaller knows within seconds
A vehicle on the crossing, a person in the track zone, sidings after hours.
Read
Parking and traffic
How many spaces are free? The camera already knows
Free spaces and congestion counted from the cameras already in place.
Read
Airports
The security queue builds at 5:40
Queues at security and check-in, the terminal forecourt, restricted zones.
ReadTry it on your own footage
Upload 30 seconds.
See what Xentim finds.
Tell us what to count or recognise and upload a short clip from a camera or a phone. In about two minutes you get it back with boxes, tracks and a counter.
The analysis is powered by GOKON v1 — our own vision model, trained specifically for outdoor events: yards, streets, car parks and squares. It recognises people, vehicles and objects in difficult, changing light — without facial identification.
- Boxes and tracks on your own footage
- A count of what you asked for
- Faces and licence plates blurred automatically
Runs on your premises, not in the cloud
Your footage stays in your building.
Xentim is a device connected to your network. It reads the video from the cameras and recorder you already have and analyses it on site. Analysis and search do not need an internet connection.
- No replacing your cameras or recorder
- No video sent outside
- No identifying people by their face
How we work
From the first call to a model
tuned to your site.
01 / Call
We listen first.
A short phone call: what your site is, which cameras you have and what the problem is today. If you like, we show Xentim on your own footage.
02 / Audit
We look for real value.
We review your cameras and how you work, and point out where the system will actually save time or filter out alarms. Where it will not pay off, we say so.
03 / Deployment
A device in your network.
We connect it to your existing cameras and set the parameters: zones, events, and who receives alerts and reports.
04 / Two weeks
We collect data from your site.
The system runs and gathers events from your cameras. This is the material we measure its accuracy on.
05 / Tuning
The model learns your site.
We analyse the collected data and detected events, then tune the model and thresholds. You get a report: how many alerts were correct before and after tuning.
Ready for questions from
your data protection officer.
Local operationFootage and analysis stay on the device
No facial recognitionA description of appearance and route, without identifying people
Face blurringAvailable in results on request
Do I need to replace my cameras or recorder?
No. Xentim connects to the video your cameras and recorder already make available on the network. If you have older analogue cameras, the recorder they are plugged into is usually enough.
Does the footage leave my company?
No. Analysis and search run on a device in your network and do not require an internet connection. Footage is never sent outside. Only the people you grant access to can see the results.
How do reports and alerts reach me?
If you want the daily report or alerts by email or messenger, the device sends them over an encrypted outbound connection. Only the report leaves: numbers, a short summary and event thumbnails, never the footage. You can also opt out and read reports on your local network only.
Does the system recognise faces?
We do not identify people. The system recognises that a person or a vehicle is in the frame and describes its appearance and route. We can blur faces and number plates in the results.
How many cameras can one device handle?
That depends on the movement in frame and the hardware chosen. We give you the exact number after the two-week trial, based on measurements on your cameras.
What is Xentim?
Xentim is an AI video analytics device that connects to the cameras and recorder you already have. It runs locally, without the cloud. You can search recordings with a plain sentence, get live alerts and receive a daily report.
Which cameras and recorders does Xentim work with?
IP cameras and recorders that make their video available on the local network, most often as an RTSP stream. The camera brand does not matter. We check compatibility with your equipment during the demo, before you buy anything.
How is Xentim different from analytics built into cameras?
In-camera analytics detect motion or line crossing in a single camera. Xentim describes every object (type, colour, route, time in frame) across all cameras at once and lets you search recordings with a sentence, for example “white van at the loading dock”. You do not need to replace your cameras.
Does Xentim help with GDPR compliance?
Xentim does not identify people and does not recognise faces, the footage stays on a device in your network, and faces and number plates can be blurred in results. You remain the controller of the CCTV data and the scope of monitoring does not change.
Will I get a lot of false alarms?
That depends on camera placement, lighting and the rules. This is why during the two-week trial we measure on your footage how many alerts the system sends and how many are correct, and then tune the model, zones and thresholds to your site.
How long does deployment take?
We start with a call and an audit of your site. Connecting the device and setting the parameters usually takes one day. Then for two weeks we collect data, tune the model and hand over a measurement report, after which you decide about permanent operation.
Can I try Xentim on my own footage?
Yes. You can upload a short clip on this page and see the analysis result, and the full demo runs on footage from your cameras.
How much does Xentim cost?
We prepare a quote after the call and the audit, because it depends on the number of cameras, the amount of movement in frame and the hardware chosen. The demo is free.
Who is behind Xentim?
Xentim is built by a team from Poznan, Poland. The product is offered by TATO STUDIO sp. z o.o., ul. 28 Czerwca 1956 r. 174/28, 61-507 Poznan, Poland.
Stop scrubbing.
Start asking.
The demo takes 45 minutes and runs on your own footage.
About us
A team that tests on itself first.
Who we are
We are a team from Poznan that has worked professionally with images for years, and for the last few with artificial intelligence models that understand video. We build and test Xentim on our own cameras before we show it to a client.
Why we do it
CCTV records everything, but almost nobody watches the recordings. When something happens, someone loses a day scrubbing through footage. We want the answer to the question of what happened to take a minute.
Book a demo
Pick a slot. 45 minutes online, shown on your own footage.
Prefer to write? contact@xentim.com