Solutions & Use Cases

No new cameras or video management system are needed, because AI Live Insight reads streams over RTSP and ONVIF from the cameras or from your existing recorder. There is no plug-in for a video management system: detections go out by webhook to any system that accepts external events, and otherwise officers work alerts on the VIDIZMO operator dashboard.

Viewing access is set per camera and per case, roles decide who can do what, every view and action is logged, and only reviewed, redacted copies leave while originals stay inside. Face recognition does nothing until the hospital enrols the people to be matched and an administrator turns it on. VIDIZMO provides data-protection, security, redaction and anonymization capabilities for PHI and PII that help customers meet HIPAA obligations, and can enter into a Business Associate Agreement where required.

VIDIZMO works closely with the hospital to calibrate the models, thresholds and rules to its own cameras and scenes, and the pilot measures the result on your footage. An officer checks the clip before responding.

Nexus and the AI models run in one on-premises deployment, in a central data center reading streams over the network or on a server at a large campus. Any NVIDIA GPU of the Ampere generation or later runs the analytics; even a consumer RTX 5090 handles about 32 to more than 100 cameras, depending on frames analysed and models per stream. You can buy direct or through a VIDIZMO partner.

Single sign-on through your identity provider, with MFA enforced there or by email passcode; role-based and per-camera access; outbound connections only to endpoints an administrator configures; shares issued per recipient with an expiry date or view limit; and export of your data through the REST API. VIDIZMO LLC is ISO/IEC 27001:2022 certified.

Not as a live detector. Behaviour outside the built-in classes is reached with prompt-driven analysis of recent frames, which runs after the fact on the language model running locally in your deployment. It does not measure a patient's condition, replace sitters or lock doors, though it can send a detection by webhook to a system that does.

They do, because streams are fetched from cameras or from any recorder or video management system that serves RTSP or ONVIF. The VIDIZMO case holds the video evidence and its custody record beside your LP case system rather than replacing it. There is no plug-in for a video management system's console: detections go out by webhook, and AI Intelligence Hub connects to any system with a REST API by configuring its endpoints.

Shoplifting is not a fixed detection class, so concealment is reached through prompt-driven analysis on the language model running locally in your deployment. That analysis runs after the fact and suits flagged periods rather than every camera all day. VIDIZMO calibrates the models, thresholds and rules for the live classes to your own cameras and stores, and the pilot measures the result on your footage.

It does not connect to point-of-sale data or correlate transactions with video, so refund and register exceptions stay with your POS exception tools.

It runs on servers you control, either centrally or at a large store or distribution center. A central server reads one stream over the network for each camera it analyses, so bandwidth follows the cameras you choose, not every camera in the store. If store recorders stay the recorders of record, the DVR window can be kept to minutes and clips around events are still kept. Any NVIDIA GPU of the Ampere generation or later runs it; even a consumer RTX 5090 covers about 32 to more than 100 cameras, depending on frames analysed and models per stream.

- Sign-in through your identity provider over SAML 2.0, OAuth 2.0 or OpenID Connect, with MFA policy staying with your provider. - Role-based access set per camera, with franchise groups or banners kept separate in one deployment. - Every clip, download and share logged, and the custody trail exportable as CSV or PDF. - Outbound connections only to endpoints an administrator configures. - Data exportable through the REST API and to your own storage. - Face recognition does nothing until you enrol people and an administrator turns it on.

Any source that serves RTSP or ONVIF can be read, so analog cameras join through the recorder that already digitises them, though analog footage gives the detector less to work with. There is no plug-in for a video management system's console: detections go out by webhook, and AI Intelligence Hub connects to any system with a REST API by configuring its endpoints. Otherwise operators work alerts on the VIDIZMO operator dashboard.

A central server reads one stream over the WAN for each camera it analyses, so the load depends on which cameras you choose. Rules can be limited to vestibules, ATMs and entrances and to scheduled hours, and if branch recorders stay the recorders of record, the DVR window can be kept to minutes while clips around events are still kept. Any NVIDIA GPU of the Ampere generation or later runs the analytics; even a consumer RTX 5090 covers about 32 to more than 100 cameras, depending on frames analysed and models per stream.

Sign-in goes through your identity provider over SAML 2.0, OAuth 2.0 or OpenID Connect, and SSO can be enforced so MFA policy stays with your provider. Access is role-based and set per camera, and regions or acquired banks can keep separate content and camera access in one deployment. Outbound connections go only to endpoints an administrator configures, and with the models running locally no model call leaves the bank. Data can be exported through the REST API.

It does not detect card skimmers or read ATM or teller transaction data, and skimmer installation is not a fixed detection class. It watches behaviour such as loitering, after-hours intrusion, robbery, fighting and weapons, and a prompt-driven review of footage for tampering can run after the fact.

It does not replace alarm monitoring: alerts go to your own operators and connected systems, and your intrusion and panic alarms stay as they are.

The platform carries no banking regulatory certification, and the attributed record of each alert, the custody trail on every clip and the account of every release support your security program's own reporting. Banks that also record contact-center calls can read about [Call Center Compliance](https://vidizmo.ai/solutions/call-center-compliance).

AI Live Insight reads any RTSP or ONVIF stream, from the cameras or from a recorder that republishes them. There is no plug-in for a video management system: detections go out by webhook, and AI Intelligence Hub connects to any system with a REST API by configuring its endpoints. Otherwise operators work alerts on the VIDIZMO operator dashboard.

The utility holds its NERC CIP obligations, and this is not a compliance product. How camera evidence supports a CIP-014 physical security plan, whether the deployment sits inside an electronic security perimeter under CIP-005, and how it is assessed as a vendor product under CIP-013, are questions your CIP team settles. For that review the system installs inside your boundary with no outbound call, sign-in goes through your identity provider with MFA, access is set by role and per camera, and outbound connections go only to endpoints an administrator configures. A reference architecture and security documentation can be requested during evaluation. VIDIZMO LLC is ISO/IEC 27001:2022 certified.

Processing can sit in a central or regional data center that reads streams over the network, or on a server at a large site, depending on how many streams each site sends and the bandwidth available. Any NVIDIA GPU of the Ampere generation or later runs it, with Blackwell preferred; even a consumer RTX 5090 handles about 32 to more than 100 cameras, depending on frames analysed and models per stream. Utilities can buy direct or through VIDIZMO's partner network.