AI Video Analytics FAQ — Deployment, Privacy & Compatibility

No. If a camera outputs an RTSP stream — which covers nearly every IP camera and NVR shipped in the last decade — Vengeea can connect to it. No new cameras, no rewiring, no changes to your existing VMS or recording setup.

1 to 3 days from kickoff to a live system, for a typical site. That covers connecting the cameras, choosing modules per zone, drawing detection zones, and setting up alert routing — no lengthy integration project, no separate ML tuning phase.

On-premises is the default architecture: video is processed locally on hardware inside your facility and never leaves your network. Cloud processing is available for teams who want zero on-site hardware, but it's opt-in, not the default — and it's encrypted end-to-end when used.

Live and running at customer sites today: intrusion, weapon detection, face recognition, fire/smoke detection, and license-plate recognition. Fight/vandalism, theft, and liquid-leak detection are in active development with early-access customers. PPE compliance, people/queue counting, and abandoned-object detection are on the roadmap and not yet built. See the modules page for the current status of each.

Tell us your facility type, camera count and the scenarios you're trying to catch. We'll size a configuration, walk you through a demo — often on your own footage — and give you a written quote. Book a demo call to start.

Vengeea is deployed at facilities across the United States, the GCC (UAE, Saudi Arabia), Europe and Kazakhstan, and is expanding into Turkey, Vietnam and Singapore. Contact us to confirm coverage and support hours for your region.

Yes — platform support, updates and tuning help after go-live, so if a zone is throwing too many false positives or a new area needs coverage, that's an ongoing conversation, not a one-time install. To be clear, this is technical support for the software, not a service where anyone watches your footage.

Vengeea works with any camera or NVR that outputs an RTSP or ONVIF stream — that covers Hikvision, Dahua, Axis, Bosch, Uniview, Milesight, Hanwha and most enterprise and prosumer brands. No firmware changes are needed, and your existing VMS keeps recording exactly as it does today; Vengeea reads the same stream, it doesn't replace it.

No. Vengeea reads the same RTSP stream your NVR already receives, so cabling, switches, VLANs and camera IPs stay exactly as they are. The Vengeea box sits on your LAN as one more consumer of a stream that's already flowing — it doesn't sit in-line with anything, so there's nothing to break if it's ever unplugged.

A single mid-range server with an NVIDIA GPU (RTX 4060 / A2000 or better) typically handles 20–40 cameras with multiple modules running in parallel — this is edge inference, so the sizing depends on camera count, resolution and how many modules are stacked per feed, not internet bandwidth. Use the server calculator to size a deployment for your exact numbers, or use Vengeea-supplied hardware if you'd rather not run your own server.

Yes. Detection, alerting, the dashboard and evidence exports all run on an air-gapped LAN, since inference happens on-site by default. Optional cloud sync — for a remote dashboard view or multi-site aggregation — can be enabled per site and stays encrypted end-to-end when it is.

Yes. Because video and biometric processing happen on-premises inside your own infrastructure by default, Vengeea operates as a data-processor with no cross-border transfer to worry about. Retention windows, deletion policies and access logs are configurable per site, which is what your DPO or compliance team will actually want to see when mapping this against GDPR, PIPL or a local biometric-privacy statute.

No one, as a matter of how the product works, not just policy. Vengeea is automated detection software, not a monitoring service — there's no operations center watching your feeds. Our staff have no standing access to your video; a temporary support session requires you to enable it, and even then it's screen-shared troubleshooting only, never a bulk export.

Face vectors are computed and matched locally against a watchlist your team builds and controls — Vengeea doesn't run against a third-party biometric database, and non-listed faces produce no persistent identity record. This matters for teams weighing biometric-privacy exposure (BIPA-style statutes and similar), not just accuracy. You can enable blur-by-default on the review dashboard and restrict identity views to named operators via role-based access.

Vengeea is subscription-based, licensed by the number of camera/module licenses you activate — billed monthly, quarterly, semi-annually or annually, whichever fits your budgeting cycle. There are no per-alert charges and no cloud egress fees. Larger deployments qualify for volume pricing and bundled module packages. Contact sales for a written quote sized to your camera count and required modules — use the calculator first if you want a technical starting point before that call.

Yes. We regularly run 2–4 week paid pilots on a subset of cameras (typically 4–12) against a written success criterion agreed up front — for example, a target reduction in false alerts against your current system, or a maximum time-to-alert on staged intrusions. Pilot fees credit against a full purchase, and the criteria are what you'd actually report up to your own leadership.

Cloud video-AI platforms require uploading footage to a third-party datacenter, which creates bandwidth cost, latency, and a compliance conversation about where your video is actually stored. Vengeea's default is on-premises: video never leaves your facility, deployment is measured in days not weeks, and there's no recurring per-GB egress bill. Cloud is available as an option for sites that genuinely don't want on-site hardware, but it's the exception, not the architecture. See the Cloud vs on-premises guide for the full comparison.

Traditional CCTV AI is usually tied to a specific camera brand or VMS and requires new hardware — AI-capable cameras or a vendor's own appliance — to get the analytics. Vengeea is vendor-agnostic: it layers on top of any camera already outputting RTSP, and modules get enabled per-camera without a truck roll or a hardware refresh cycle. See the Vengeea vs Traditional CCTV AI page for the full comparison.

Logistics warehouses, industrial and chemical plants, energy substations, ports and terminals, retail, conference & event venues, and public-sector critical infrastructure. The modules most commonly enabled by industry: intrusion and LPR for logistics and ports, weapon and face match for events and government sites, fire and smoke for industrial and chemical facilities, and theft detection for retail (in early access).

Yes. Vengeea supports customer-supplied detection models (YOLO-family, ONNX) that run alongside the shipped modules on the same feeds. If you don't have a model, our team can scope and build a detector for a specific object or behaviour under a scoped engagement — request a custom module.