VisualOps connects to the cameras you already run. Type what you are looking for and get the clip in seconds. Set rules once and get an alert the moment something needs a person.
Hundreds of cameras, a small security team, and hours of scrubbing to answer one complaint.
You know when a car came in and when it left, but not where the hours in between went.
Cars move between showroom, yard, test drives and workshop, and nobody has one view of where each one is.
Public halls get crowded, complaints arrive weeks later, and finding the footage takes a full shift.
A vision-language model indexes every camera continuously. Ask for a person, a vehicle, an action or a situation, with a place and a time if you know them. VisualOps returns ranked clips with the camera, timestamp and a still of the match. No tagging, no pre-set categories.
Draw a zone, pick a condition, choose who gets told. Alerts arrive on WhatsApp, SMS, email or your control room screen with a 10-second clip attached, so the person on duty can decide without opening another system.
The same search and alert engine, set up for very different days.
Large sites with hundreds of cameras and a small security team. VisualOps watches the perimeter and gates overnight and gives investigators answers in minutes, not shifts.
Floor teams already have cameras. VisualOps turns them into a daily read on service: how long customers wait, whether counters and bays are working, and where every vehicle is right now.
VisualOps runs on an edge appliance inside your network. Video is processed where it is recorded, and only events, thumbnails and the search index leave the site if you choose.
The query is parsed into what to look for, where and when. Attributes like colour, clothing, vehicle type and action are matched against the embedding index. The top candidates are then checked frame by frame by the language model before they are shown, which removes most near-misses.
Search works in English and Hindi today. Results respect user permissions, so a store manager only sees their own store.
ANPR reads standard, BH-series, commercial and two-wheeler plates, including dirty and partially covered ones, at speeds up to 40 km/h at gates. Reads are matched against your watchlists and visitor passes.
Crowd density counts heads per zone rather than detecting whole bodies, so it holds up in packed OPD queues and festival processions where people overlap.
Rules combine a zone, a condition, a schedule and a cooldown. Every alert carries a clip and a still, and the operator can mark it as resolved or false, which is used to tune thresholds for that camera.
Add units to scale. Hybrid and cloud options are available for distributed retail.
| Streams per unit | Up to 32 at 1080p, or 64 at 720p |
|---|---|
| Form factor | 2U rack mount, or compact unit for single stores (8 streams) |
| Inputs | RTSP, ONVIF Profile S and T, VMS connectors |
| Index retention | 90 days default, configurable up to 1 year |
| Network | Works offline; outbound HTTPS only for alerts and updates |
| Time to first alert | Same day as installation |
| Data location | On your premises. Cloud components hosted in India. |
|---|---|
| Face handling | No face recognition by default. Optional face blurring in exported clips. |
| Access | Role and site-based permissions, SSO with SAML or OIDC |
| Audit | Every search and export is logged with user and reason |
| Compliance | Designed for DPDP Act 2023 obligations |
VisualOps sits next to your NVR or VMS, not in place of it. Alerts go where your team already works.
Any camera that streams over RTSP or ONVIF. Analog cameras work through their DVR.
Read streams from the VMS and push events back as bookmarks on its timeline.
Link a door event or alarm to the camera that saw it, and verify alarms before guards are sent.
Each rule chooses its own channel and escalation path.
Match till transactions and job cards to video, so every bill or repair order has the footage behind it.
Feed events into command centres and ticketing systems.
Every alert can post to your endpoint. The REST API exposes search, rules, cameras and clips, with keys scoped per site.
POST https://your-system.example/visualops-events { "event": "anpr.watchlist_match", "site": "Main campus", "camera": "Gate 2 · Inbound", "plate": "MH12KX4471", "confidence": 0.97, "clip_url": "https://edge-01.local/clips/8f3a2c", "occurred_at": "2026-10-06T02:14:33+05:30" }
Four deployments across healthcare, a city command centre, a pre-owned car retailer and a grocery chain. Client names are withheld at their request.
Nine clinical blocks, a medical college, staff housing and multi-level parking, with 6,000 to 8,000 people on campus on a weekday. The security team of 40 had 640 cameras and no practical way to use them beyond live viewing.
A tier-2 city's Integrated Command and Control Centre had cameras at 260 junctions and 40 public spaces. Operators watched video walls, but investigations still depended on manual review.
"two men on a black scooter without helmets near the old bus stand, Saturday night"
Returned 7 clips from 3 cameras across 2 days of footage in under 5 seconds.
The retailer buys cars from individual owners, refurbishes them at regional hubs and sells them online with a warranty. Each hub holds 400 to 700 cars, moving between inspection, mechanical work, body and paint, detailing and the photo studio. Stage updates were entered by hand in an app, often hours late, and sales teams could not tell a buyer when a car would be ready.
| Plate | Car | Stage | Time in stage | Last seen |
|---|---|---|---|---|
| MH12 RT 8841 | 2019 Hyundai Creta SX | Body and paint | 31 h plan 24 h | Paint booth 2 · 4 min ago |
| KA03 MN 1207 | 2018 Maruti Baleno Zeta | Mechanical | 5 h 40 m | Bay 6 · on lift |
| MH14 GP 5530 | 2020 Honda City VX | Detailing | 2 h 10 m | Wash bay 1 · 1 min ago |
| DL8C AX 9012 | 2017 Toyota Innova Crysta | Inspection | 48 m | Inspection pit · now |
| MH12 SK 3398 | 2021 Tata Nexon XZ+ | Ready for sale | 1 d 3 h | Yard row F, slot 14 |
| MH04 JL 7765 | 2019 Kia Seltos HTX | Test drive | 22 m | Exited Gate 1 · 11:42 |
Long checkout queues at peak hours were the top complaint in customer feedback. Store managers could not watch the tills and run the floor at the same time.
30 minutes with a solutions engineer. Bring the questions your team asks about video today.
We will run VisualOps on footage from your sector and set up alerts while you watch.