Security for vacant property should end when the vacancy ends. Here is the device that does.

Because vacancy is a finite event and subscriptions are not. A flip sells in 30 to 180 days. A probate estate closes in 90 to 240 days. Bank REO inventory rotates in 6 to 18 months. Receivership varies. In every case, the security expense should end when the vacancy ends, otherwise you pay for an empty service that is either auto-renewing unmanned camera feeds or worse, auto-renewing on a property that is now occupied and shouldn't be watched at all. The incumbents (Pro-Vigil, Stealth, SentriForce, DAWGS, Clearway, Ajax, EyeQ) price this as recurring revenue because it is structurally good for them, not because it matches the customer's exposure window. A device you own, install, and unplug matches the exposure window exactly.

One unit can cover up to 25 cameras per property. The per-property marginal cost after the unit is paid off is zero hardware and roughly $200 per month of active monitoring (the subscription for cloud control plane, WhatsApp routing, and event storage), billable only for the months the unit is plugged in and configured against a site. A flipper holding three properties a year for 90 days each amortizes one unit across the portfolio: nine months of active monitoring per year, three months of quiet ownership. Compare to a per-property Pro-Vigil or Stealth contract billed for twelve months whether the property is occupied or not.

At install, you unplug the HDMI cable from the DVR to the guard monitor, plug the DVR side into the Cyrano unit's HDMI in, plug the Cyrano unit's HDMI out into the monitor. The unit acts as a transparent passthrough, so the guard monitor keeps showing the same multiview it always did. The unit never logs into the DVR, never talks on the DVR's LAN, and never changes the DVR's user accounts, password, recording settings, or RTSP config. At uninstall, you reverse the cable swap. The DVR is bit-identical to before install because the unit never touched its management plane. This matters when a vacant property sells and the buyer runs their own audit on the existing surveillance stack.

On the unit: the layout_id cache (remembers the DVR's specific 4x4 or 5x5 HDMI compositing mode), the tile-to-camera map (which tile on the grid is lobby, which is rear entry, etc.), the zone rules per tile (whitelist hours, dwell thresholds), and the WhatsApp or SMS routing for this property's alerts. Off the unit, in the cloud control plane: event history, clip storage, portfolio dashboard. When you physically relocate the unit to the next vacancy, you factory-reset the on-device state through a short provisioning flow (about 2 minutes), and the previous property's events remain in the cloud for the record. The ops bundle is portable because the sensitive binding between unit and property is narrow and local.

A 4G or 5G cellular modem is enough because the unit does not stream video anywhere. Raw multiview stays on the DVR, inference runs on the unit at the property, and only event packets (a bounding-box class, a timestamp, a ~220 KB clip, a thumbnail) leave the site. That is typically a few megabytes per day even on an active site, which fits inside any prepaid SIM plan. This is the default state of vacant property: power still connected because the meter has not been pulled, internet canceled because the business account closed with the last tenant. Incumbent services that stream RTSP continuously to a SOC need sustained uplink and fail silently when the prepaid plan data cap is hit.

Two options. Option A, cheaper for you: you unplug the unit and move it to your next vacant property, and the new owner reinstates their own surveillance contract (which most do anyway). The DVR is left in the exact state they bought it in because the unit never touched it. Option B, rare: you sell the unit with the property, and the new owner re-provisions the device against their own WhatsApp routing. The in-unit state wipes on the factory reset at the start of re-provisioning. Both options avoid the incumbent pattern where a subscription silently transfers to the new ownership and nobody notices until a billing dispute surfaces months later.

Yes, because the install surface is one HDMI cable and one cellular modem plug, both already standard on any vacant site that had a running DVR. The sequence is: unplug the monitor-side HDMI cable from the DVR, plug the Cyrano unit between the DVR and the monitor, connect the cellular modem, power on. The longer step is configuration, which is done once per new site over a short provisioning flow that maps tiles to camera labels by looking at the DVR's live multiview and naming each rectangle. On an 8 or 16-camera site, the full hands-on time is 5 to 10 minutes. The physical swap is the 2-minute part.

The unit is rack-mounted alongside the DVR in the same closet or equipment room that already houses the existing surveillance hardware. If an intruder can reach the DVR, they can reach the unit. However, the device has no local admin interface accessible without physical provisioning and does not store raw video, so a theft of the unit does not expose footage. The DVR's own disks remain on site, and the cloud control plane retains all alert history. The practical exposure from a stolen unit is the hardware cost, not a privacy leak. For higher-value sites, a lockbox around the DVR-and-unit pair is standard practice.

The DVR keeps recording to its own disks throughout, the way it always did. The Cyrano unit adds detection events and short clips on top, routed into WhatsApp or SMS for fast response. For an insurance claim or police report, the flow is: the alert tells the receiver or asset manager exactly when and where something happened (tile=12, zone=rear_entry, 03:14 AM), and they export the DVR's native footage for that timestamp and submit that as the evidence. The AI is the index into 24 hours of passive video, which is the part that used to take a day of manual scrubbing.

Because the capture point is HDMI, not RTSP. Every DVR in the last decade outputs a composite multiview on HDMI, regardless of whether the cameras feeding it are analog BNC from 2012 or PoE IP from 2024, or a mix. The unit reads the composite, not the individual cameras, so camera brand and protocol do not matter. What changes per property is the layout_id (4x4 vs 5x5 vs 3x3) and the tile-to-camera map, both of which are set during the 2-minute provisioning step. That means the same unit genuinely works at a Fort Worth Class C apartment with 16 analog cameras, then a shuttered retail strip with 24 IP cameras, then a probate estate with 8 cameras. The input surface is identical: one HDMI cable.