Your control room's smartest operator. Ask in English, Hindi or Hinglish, and Radha finds the footage, follows the vehicle, moves the camera, writes the rule and briefs the next shift. All on your own servers.
Radha is the only VMS copilot we know of that is conversational, takes action and runs fully on premise. She commands 111 tools across the whole of Visylix: plain-language video search, investigations with evidence timelines, exact plate journeys and person journeys, heat-map hotspots, PTZ presets and tours by name, face enrolment, false-alarm marking, automation authoring and shift briefings. Thirteen high-impact actions always ask first, and every call is written to the audit trail.
111
Tools
13
Actions That Always Ask First
3
Languages: English, Hindi, Hinglish
0
Cloud Dependencies
Most assistants answer questions. Radha gets the work done: she runs operations, investigates incidents, writes reports and learns your site. Every part of Visylix can be run through conversation.
Ask "show me every face detection from the car park cameras between 2pm and 4pm yesterday" and Radha does it straight away. She answers in English, Hindi and Hinglish.
Radha runs on an on-premise language model on your own servers, with scene understanding and speech recognition beside it. Every conversation, image and voice clip stays inside your building, which is exactly what government, defence and banking teams need.
Tell Radha once to "call this camera the gate cam" and she remembers. She keeps each user's preferences and your team's corrections, all stored on your own hardware.
Ask an open question and Radha works it like an investigator, checking sightings, detections, events and alerts, then hands you findings with an evidence timeline you can replay moment by moment. Fully air-gapped.
Important alerts arrive with a plain-language note on what happened and why it fired, so operators can clear the alert rail quickly and confidently.
Radha has no delete tool. Her 13 high-impact actions never run on her word alone: she names exactly what will change and waits for your click, and she cannot approve her own actions.
From one person's movements to a full incident investigation, in English, Hindi or Hinglish.
When was James recognised today?
James was recognised 4 times today: 08:52 at the Main Entrance, 09:14 in the Lobby, 13:03 at the Cafeteria camera, and last at 17:41 leaving through the Employee Gate. Shall I play any of these moments?
Find a person in a red jacket near the entrance this morning
I found 3 matches this morning: 08:12 at Main Entrance (closest match), 08:47 at Side Entrance A, and 11:20 at Parking Entry. Shall I open the 08:12 recording at that exact moment?
Loading dock pe aaj dopahar kuch unusual hua kya?
Maine 12:00 se 18:00 tak loading dock check kiya: 14 person detections, jo weekday ke liye normal hai; 14:22 par ek line crossing, scheduled delivery ke dauraan; aur 16:48 par 6 minute ki loitering jo kisi delivery window se match nahi karti. Timeline ready hai. 16:48 wala event dekhna chahiye. Play karoon?
Brief me on my shift and make a printable report
Your shift: 24 of 24 cameras online, 7 AI alerts (2 unacknowledged, both low severity), no PPE violations, storage at 61% with 34 days to spare. Nothing needs escalation. Your printable report is ready. Shall I send you the daily digest by email as well?
Radha does the real thing on your live system, within your permissions. Thirteen high-impact actions always ask first, every call is audit-logged, and nothing can be deleted.
Open any camera, start or stop streams and recordings, take snapshots, control playback and nudge PTZ cameras, just by asking.
"Show me the gate camera live"
"Pan camera 4 left and zoom in"
Ask about faces, number plates, fire and smoke, weapons, intrusions, PPE, crowds, sounds and more, with dates and cameras. See heat-map hotspots, and mark a false alarm in one sentence.
"When was James recognised today?"
"That fire alert on camera 6 was a false alarm"
Describe what something looks like and Radha finds it in your footage, then pulls sightings, detections, events and alerts into findings with a time-ordered evidence timeline.
"Find a man in a red jacket near the entrance"
"What happened at the dock this afternoon?"
Set up a live watch or a full Sudarshan rule from one sentence, then test it against past footage before it goes live. New rules start as drafts, and switching one on always asks for your confirmation.
"Watch for plate MH12* tonight"
"Test that rule against last week"
Ask why a camera is down and Radha checks it step by step: network, password, stream and recent errors, then tells you what to try. Apply the fix to many cameras in one confirmed step.
"Why is the dock camera offline?"
"Restart every camera in the warehouse group"
Ask how many more cameras a server can take and Radha measures its real headroom, with separate answers for video only and for cameras running AI.
"How many more cameras can I add?"
"How many with face recognition on each?"
Reach footage stored on your existing NVRs. Radha lists your recorders, searches their archive by time and plays the clip right in the chat.
"What does the NVR have for gate 2 on Tuesday?"
"Play that back from 14:30"
Send a camera to a named preset, or start, stop or check a patrol tour, without hunting through the PTZ panel.
"Send the yard camera to the loading bay preset"
"Start the overnight patrol tour"
Work the alarm console by conversation: open an alarm, read the detail, move it through triage, check sensor history and play a talkdown warning through the camera speaker.
"What is on the alarm console right now?"
"Play the trespass warning on the yard camera"
Jump to the exact moment on any camera, bookmark recordings, search the audit trail and lock footage as evidence, sealed with SHA-256. Nothing can be deleted through Radha.
"Show me camera 2 at 14:32 yesterday"
"Lock recording 42 as evidence"
A spoken briefing of cameras, alerts, PPE and storage, on demand or as a daily digest by email. One sentence becomes a structured, printable incident report.
"Brief me on my shift"
"Make a report of that fire alert"
Add cameras, scan the network for ONVIF devices, create users, limit who sees which camera, enrol faces and plates, manage sites, groups and floor plans, test integrations, activate licences and restart services.
"Enrol James from snapshot 12"
"Let Priya see only the lobby cameras"
Radha answers questions about your own Visylix system, such as version and what your licence covers, and remembers your site's names and preferences.
"What does my licence cover?"
"Call this one the gate cam from now on"
Built for organisations where privacy is not optional, and where an AI with real power needs firm guardrails.
Language, scene understanding and speech all run on your hardware. No outside services, no telemetry and no data leaving your network. Radha works in completely air-gapped environments.
Arming a rule, restarting services or changing many cameras at once: Radha proposes the action, names the exact target and waits for your click. She cannot confirm her own actions.
Radha follows each user's role, camera access and licence on every action, and writes every action to the tamper-evident audit trail. Nobody can do through Radha what they could not do themselves.
Searches based on race, ethnicity or religion are refused. Attempts to trick Radha with hidden instructions are blocked and logged. She answers only about your video system, never the internet.
No other VMS we know of has a copilot that is conversational, takes action and runs entirely on premise. Cloud vendors keep their AI in their cloud; on-premise vendors offer a search box at most. Radha talks, searches, investigates, sees, listens and runs your system, with no cloud at all.
Every request is carried out for real, within your permissions. Ask Radha to enrol a person, limit a user to certain cameras or lock a recording as evidence, and she does it and confirms the result.
Radha investigates. Ask "what happened at the dock this afternoon?" and she works through sightings, detections, events and alerts to give you findings and an evidence timeline you can replay.
Radha speaks your team's language: English, Hindi or Hinglish, with your own camera nicknames rather than system IDs.
Radha is governed. No delete tool, confirmation on 13 high-impact actions, refusal of discriminatory searches, protection against hidden instructions, role and licence checks on every call, and a full audit trail.
Book a live demo and see what it is like to run your whole control room through conversation.
Radha is the AI copilot built into Visylix. Operators ask for what they need in plain words, in English, Hindi or Hinglish, and Radha does it: finds footage, traces a vehicle or person across cameras, moves a PTZ camera, writes an automation, briefs the next shift or onboards a camera. She runs on an on-premise language model, so every conversation stays inside your building. Radha is the only VMS copilot we know of that is conversational, takes action and runs fully on premise.
Radha commands 111 tools that cover the whole of Visylix: live view, recordings, snapshots and playback; PTZ presets and patrol tours by name; answers from every AI analytic you run; exact plate journeys and appearance-based person journeys; heat-map hotspots; plain-language video search and multi-step investigations; Sudarshan automation rules with a dry run before they go live; the alarm console, sensors, notifications and talkdown; shift briefings and incident reports; evidence locks; camera health checks and bulk actions; and administration from camera onboarding and user management to face and plate enrolment, zones, sites and floor plans. Radha has no delete tool, so nothing can be removed through her.
Yes. Ask "find a man in a red jacket near the entrance" or "was there a white van here this morning?" and Radha returns time-stamped matches you can play straight away. The search runs on your own servers, and no footage or question ever leaves your network.
Yes. Ask "what happened at the loading dock this afternoon?" and Radha works through people sightings, AI detections, system events, alerts and video search, then gives you her findings with a time-ordered evidence timeline. She can turn the result into a printable incident report.
Radha answers in English, Hindi and Hinglish, so operators can talk to her the way they talk to each other. The Visylix interface itself is available in 55 languages.
Yes. Radha can look at any live camera or snapshot and describe the scene ("what is happening on the office camera right now?"), and she accepts spoken questions. Both run on your own hardware; no cloud vision or speech service is ever used.
Yes. Tell Radha an alert was a false alarm and she marks it for you. On PPE, object, fire and weapon alerts, that camera's threshold then tightens automatically, never below the limit your operators set.
Yes. Tell her once to "call this camera the gate cam" and she understands the nickname from then on. She remembers each user's preferences and corrections, and all of it is stored on your own hardware.
No. Radha runs entirely on premise, including the language model, scene understanding and speech recognition. She needs no internet access and works in fully air-gapped sites, which makes her a natural fit for government, defence and other high-security networks.
Radha follows every user's role, camera access and licence on every action, and every call she makes is written to the audit trail. She has no delete tool. Thirteen high-impact actions, such as arming rules, restarting services or bulk camera changes, always ask the user to confirm the exact target first, and Radha cannot approve her own actions. Searches based on race, ethnicity or religion are refused, and attempts to trick her with hidden instructions are blocked and logged.
Most VMS assistants either run in the vendor's cloud or stop at a search box. Radha is a full copilot that talks, searches, investigates, sees, listens, reports and takes action, all on your own hardware, in English, Hindi and Hinglish, with confirmation on high-impact actions and a complete audit trail.