AI video analytics you can talk to.
Ask a question and get the answer from your CCTV. Microzeit summarizes hours of video and finds any person or vehicle by face, clothing, colour or number plate, powered by our NIST-ranked recognition engine.
Demo of the Microzeit AI assistant on sample footage. It summarizes a clip from a security gate into four key moments, finds people wearing a blue jacket on a street camera, finds yellow taxis at a junction, and finds where a person was seen from a photo of their face.
Deployed with partners across India
Ask your cameras anything.
Our in-house vision language model describes every frame in plain English. Summarize a day of footage in minutes, ask it a question, search for any word, set alerts in plain language and read signboards in Hindi, Marathi and English as the video plays.
- Video summarizationSummarize hours of video in minutes.
- Video forensicsBuild a case and find the moment that matters.
- AI video searchAsk your footage a question.
- Signboard reading & live translationRead signboards and translate them as the video plays.
- Custom AI alertsDescribe what to watch for. Get an alert when it happens.
Every summary line links back to its frame, and the assistant answers from those summaries.
One platform for every video analytic.
Pick the analytics each camera needs. They all run on the same cameras and the same engine, so adding one is a setting, not a new project.

Video intelligence
Our in-house vision language model reads every frame, so hours of footage become text you can read, search and set alerts on.

Security & perimeter
Intrusions, threats, tampered cameras and incidents, flagged the moment they happen.

People & crowd analytics
Counts, crowd density, movement and demographics across stations, campuses and public spaces.

Traffic & vehicles
Number plates, vehicle types and traffic flow at every junction and gate.

Workplace & sites
Safety gear and attendance on factory floors, offices and campuses.
Powered by recognition ranked by NIST.
Every analytic runs on our own face, body and vehicle recognition. First in class in NIST's facial recognition evaluation. Across seven independent trials. In three of them, the algorithm set a record in the entire history of testing.
See every result since 2015Results recognized by
- NIST
- IARPA
- WIDER FACE
- ActEV · CVPR
- EmotioNet
- Kaggle
- NISTFirst in class across seven independent trials, with records in three
- IARPA1st in verification accuracy and identification speed
- ActEV · CVPR2nd in activity detection in extended video
- WIDER Challenge3rd in pedestrian and cyclist detection on public VMS footage
What it detects on your cameras
Built for camera networks that never switch off.
- Live and recorded, one platform
- Real-time alerts on live cameras, and summaries and search across everything you've recorded.
- On the cameras you have
- Works with your existing CCTV over RTSP, on your own servers or in the cloud.
- Ranked by NIST
- The face recognition underneath was first in class across seven independent trials.
- 0.3 s across 1.5 B faces
- Search a database of 1.5 billion facial images in 0.3 seconds, on an unlimited number of streams, 24×7.
Where it's used.
Choose how you apply recognition to suit your security needs, your policy on personal information and the devices you have. If your use case isn't covered out of the box, we build it.

General security
Spot persons of interest across every camera on your premises.

Access control
Let staff and visitors through by face, with liveness checks at the gate.

Retail
Understand footfall and visitor mix, and get alerts for known shoplifters.

Visitor identification
Register visitors once and recognize them every time they return.

eKYC
Verify a customer's face against their ID photo, remotely.

Safe city
Watchlists, crowds and incidents across a city's camera network.

VIP & celebrity identification
Recognize VIP guests the moment they arrive.

Automatic attendance
Mark attendance by face as people walk in.

Drone surveillance
Run recognition on aerial footage from drones.
Not on the list? Tell us your use case. If it isn't available out of the box, we build it. Message us
Runs where your footage lives.
- On-premises or cloud
- Install it on your own systems, run it in the cloud, or both.
- Works with your software
- An extensive set of REST APIs connects it to your video analytics application and other third-party software.
- Your cameras stay yours
- We provide the software, support, and camera and hardware sizing. The cameras and hardware remain your own.
# Search a photo against a watchlist
curl -X POST https://api.example/v1/faces/search \
-H "Authorization: Bearer $API_KEY" \
-F "image=@visitor.jpg" \
-F "watchlist=enrolled-staff"
# Response
{
"matches": [{ "id": "0042", "score": 0.97, "camera": "gate-2" }]
}Our journey.
2015
Company founded
Microzeit was founded in Bangalore as an IT consulting company, delivering, developing and managing large-scale IT projects for Aurea.
2020
Facial recognition and video analytics
Microzeit began building facial recognition and video analytics solutions.
Core values.
- Innovation
- We build our own models, including the vision language model behind video summarization and search.
- Transparency
- The system says what it can't see: plates marked not legible, workers marked unverified, answers linked to their frames.
- Accessibility
- Recognition that runs on the cameras you already have, in the cloud or on your own servers.


























































