A Genetec Security Center plugin that lets operators find any clip by simply describing what they're looking for — no timeline scrubbing, no hardcoded analytics categories.
Genetec Security Center is the platform of choice for serious security operations — but when an incident happens, finding the relevant footage still means manually scrubbing through hours of archive. Operators work backwards from a timestamp, hunt across camera channels, and spend the bulk of their investigation doing nothing but watching empty corridors at 4×.
There is no native way to ask Security Center a question. You cannot type "show me the delivery van that came through the service entrance after 7 PM" and get a result. You can only filter by bookmark text you thought to write ahead of time, or metadata from analytics engines that have to be configured per-camera, per-event, long before the incident occurs.
Xplug.Vision closes that gap. It continuously watches your cameras, builds a searchable index of what's happening in every frame, and lets any operator ask it a question in plain English — without touching a timeline.
Type a description of what you're looking for — a person, a vehicle, a behaviour, a scene — and get back a ranked list of matching clips with thumbnails, timestamps, camera names, and confidence scores. No special syntax, no training required.
A background service continuously pulls keyframes from every configured camera, uses a vision model to generate a rich description of each frame, and stores the result as a searchable vector embedding. By the time an incident is reported, the footage is already indexed.
Xplug.Vision is a native Workspace SDK plugin — it appears as a panel inside the operator's existing Security Desk environment. Clicking a search result jumps the active tile directly to that moment in the archive. No new application to open, no context switching.
Designed from the ground up for large deployments. Hundreds of cameras, multiple concurrent operators, 24/7 indexing — all running against a self-hosted vector database that keeps your footage data inside your own infrastructure.
Every query is logged — who searched, when, what they typed, and which results they clicked through to. Security operations need accountability, and Xplug.Vision provides a complete record of every investigation that touched the system.
Administrators choose which cameras are indexed and at what frequency. Sensitive areas can be excluded entirely. Indexing runs as a Windows service or Docker container and connects directly to Genetec's Media SDK for frame extraction.
A background indexing service connects to Genetec's Media SDK and pulls keyframes from every configured camera on a rolling schedule. Each frame is passed to a vision model that generates a detailed plain-English description of the scene — people, vehicles, behaviour, lighting conditions, notable objects.
Each generated description is converted into a vector embedding and stored in a PostgreSQL database with the pgvector extension. An HNSW index ensures that similarity searches across millions of frames return in milliseconds, even on modest hardware.
The operator types a natural language description directly into the search panel inside Security Desk. The query is converted into a vector embedding using the same model used for indexing, then compared against every frame in the database to find the closest semantic matches.
The top matching clips appear in the panel ranked by confidence score, each showing a thumbnail, the camera name, and the exact timestamp. Clicking any result seeks the active video tile to that precise moment in the archive. The search and the click-through are both logged to the audit trail.
Xplug.Vision is a Genetec Workspace SDK plugin — it installs the same way every other third-party Security Desk extension does. There is no cloud dependency for your footage. Frames are extracted and indexed on-premises, and the vector database runs inside your own infrastructure as a Docker container or a dedicated Windows service.
The only external API calls are to the AI services used for frame description and embedding generation. Those calls contain only the image frames and text — no camera identifiers, no site metadata, no personally identifiable information.
Discuss deploymentWe're working with a small group of Genetec partners and end-users during Phase 1. If you're running Security Center and want to see Xplug.Vision in your environment, get in touch.