Every home-monitoring page sells the same panel plus contract. This one is about the 8-channel DVR already on your shelf and how it becomes the monitoring system.
The entire first page of Google for "security systems home monitoring" reads as a product catalog. ADT, SimpliSafe, Vivint, Ring Alarm, Cove, Frontpoint, abode: different branding, same bundle. A proprietary panel. A short list of door contacts and PIRs. An optional cloud camera or two with a per-device storage fee. A $15 to $50 monthly bill that pays a UL-listed central station to answer the phone when the panel trips.
That is not the only shape of home monitoring. A large number of homeowners already have a different setup: a four, eight, or sixteen channel NVR from Costco or Best Buy or Amazon, coax or PoE runs to the four corners of the house, a monitor in the garage, and no contract because they only wanted recording. The monitor shows the grid. Nobody is watching it. This guide is about the move that turns that grid into a real-time monitored system, without replacing the cameras, without replacing the recorder, and without paying a central station.
What the top five home-monitoring pages all miss
Read Security.org, SafeHome.org, SafeWise, Defender, and the dozen SEO blogs ranking around them. Every single guide carves the decision the same way: pick your kit (ADT, SimpliSafe, Vivint, Ring Alarm, Cove, Frontpoint, abode, Brinks), pick your monitoring tier (self-monitored app notifications or professional 24/7 monitoring), and pick your optional cameras. The camera portion is always an add-on, priced per-camera, with its own cloud storage tier.
None of them describe the common case of a homeowner who already solved the camera portion years ago. A Lorex 4K eight-channel NVR kit on sale at Costco covers four or eight exterior angles for under $700. A Reolink PoE NVR with six cameras is a weekend install. The recorder has HDMI-out, plays a multiview on a 24-inch monitor in a closet, and writes footage to an internal drive for thirty days. That homeowner reads a guide that tells them the way to "monitor" their home is to rip out what they have, buy a new kit, and start paying someone $45 per month. The gap is that the guide and the homeowner are optimizing for different things.
Monitoring is not "owning the hardware from the people you pay." Monitoring is "a classifier plus an alerter plus a human in the loop." The recorder in the closet already has the cameras and already has the feed. What is missing is the classifier. The rest of this guide is about how a small box tapped into HDMI-out fills that gap.
Two paths to a 2 AM alert on the back porch
Same event, two pipelines. In the first, the industry-standard panel contract. In the second, the HDMI edge-tap. The difference is not speed; it is how many actors have to touch the signal before the homeowner sees anything.
PATH A: PANEL + CENTRAL STATION
PATH B: HDMI EDGE-TAP ON EXISTING NVR
Path A has seven message hops because panel sensors tell you "something tripped," and a human has to decide what. Path B has four because the decision happens on the device before the first outbound message. The thumbnail is the verification. Nothing else has to happen.
The fact this page exists because of
The whole approach only works because consumer NVR recorders all paint their multiview onto HDMI-out in similar enough ways that a single template library can recognize the chrome across brands. That library is the uncopyable middle layer. Here is what is covered today and what gets blanked before inference.
Overlay classes blanked
The recorder draws a clock, a camera-name strip along the bottom of each tile, and a channel bug in the corner. All three are painted onto the same HDMI output the cameras use. Inference sees them as text, not objects, which is expensive noise. The overlay_mask blanks clock, cam_name_strip, and channel_bug before a single bounding box is drawn.
Home-grade NVR brands covered
Lorex, Amcrest, Reolink NVR, Swann, Night Owl, Q-See, ANNKE, and EZVIZ. These are the brands on the Costco and Best Buy shelf. All of them render a 2x2, 3x3, or 4x4 multiview with the same general chrome. All of them are in the template library out of the box.
Commercial brands covered too
Hikvision DS-7xxx, Dahua XVR and NVR, Uniview, Bosch DIVAR, Honeywell Performance, and Panasonic WJ-NX. Same overlay_mask machinery, different brand strings. A mixed-brand portfolio is fine.
Layouts recognized
1x1 (single camera), 2x2-std (four cameras, the most common home setup), 3x3-std (nine cameras), 4x4-std (sixteen, the ceiling for most home NVRs), and 5x5-std (twenty-five, the Cyrano per-unit maximum). The layout_id is emitted with each event so the tile.coords are interpretable.
Why the HDMI tap is the right interface for this
The obvious-seeming alternative is to pull the camera streams directly, either over ONVIF or by hitting whichever proprietary RTSP URL the NVR exposes. Every homeowner who has tried this knows how it ends. Three of the eight cameras speak ONVIF, two speak a custom RTSP with firmware-specific quirks, and one requires you to enable a hidden debug menu to find the stream URL. The HDMI-out avoids all of that.
Pulling streams vs tapping HDMI-out
Per-camera credentials, per-brand quirks, firmware drift, and every new camera you add is a new integration task. You end up maintaining a small fleet of brand-specific decoders.
- Different URL per camera brand
- Auth tokens rotate on firmware updates
- New camera = new integration
- Breaks when NVR firmware updates
What the classifier has to do before an event leaves the house
The detector runs continuously. Raw object detections come off it at roughly 412,803 per 24 hours on a 25-tile multiview. Nothing you want to see is in that raw stream; it is drowning in wind-swayed branches, a neighbor's headlights, a fly crawling across a lens, and the recorder's own clock re-drawing once per second. Four filter stages run before any alert crosses the network.
Local pipeline
The thing that makes those numbers usable is that 0% of candidate detections never cross the network. The filters run on the device in your closet. Whatever gets through has already been verified by four independent checks.
The exact shape of the event that lands at 2:14:07 AM
The homeowner does not need to know what is inside the envelope, but the homeowner also deserves to. Every event carries the same nine fields plus a single thumbnail. The fields exist because each of them answers a question the homeowner will ask before acting.
Nine fields, nine reasons
- tile.label — the camera name the NVR itself stamps on the strip, so 'Back Porch' is the recorder's own text, not a rename in the cloud.
- tile.index — which slot of the grid, so the render layer knows where to draw the highlight.
- tile.coords — pixel box inside the composite frame, for the 480x270 crop.
- property — which house or site. A homeowner with a rental and a primary gets the property name on every event.
- layout_id — 2x2-std for four cameras, 4x4-std for sixteen. The coords only make sense with the layout.
- overlay_mask — which chrome classes were blanked before inference, so the detection is reproducible.
- event_class — person_in_zone, vehicle_dwell, loiter, or tamper. This is the one word the homeowner reads first.
- iso8601_ts — the recorder's clock, so the deep link lands on the exact second in the archive.
- latency_ms — capture to delivery, so you can see when a slow upstream delayed the notification.
What installation actually looks like
The entire install is a cable swap. The monitor that currently plugs into the NVR's HDMI-out now plugs into the HDMI-out of the Cyrano device. The Cyrano device's HDMI-in goes into the NVR. That is the install.
Physical install, one afternoon
Find the NVR's HDMI-out
On a Lorex, Amcrest, Reolink, Swann, or Night Owl, it is on the back of the recorder and currently feeds a monitor in a closet or garage. Label it before unplugging.
Unplug the monitor from the NVR
The NVR keeps recording. Cameras keep rolling. Disk keeps filling. Nothing about the camera side changes.
NVR-out into Cyrano-in
HDMI cable from the back of the NVR into the HDMI-in port on the Cyrano device.
Cyrano-out into the monitor
The same monitor that was there before now plugs into the HDMI-out port on the Cyrano device. The grid looks identical because the device is passing the signal through.
Network the Cyrano device
Either plug in ethernet or configure Wi-Fi. Detection starts running on-device immediately; events only stop being local-only once the uplink connects.
Draw per-tile zones
In the dashboard, click the camera tile and draw the polygon of the zone you care about (the back porch, the driveway, the side gate). Zones are where the tile-grid gate filter applies.
Live on the device, seen on the wire
If you prefer to look at the system from the outside rather than the dashboard, this is what the first five minutes of a fresh install looks like from the command line. Detection runs before the uplink, and the queue drains once the network comes up.
Cost per month compared honestly
The traditional stack is a mortgage of small monthly fees. The HDMI edge-tap is a one-time device cost plus a single monthly number that does not scale with camera count. Here is the comparison at a typical four-camera home and at a small landlord with four properties of four cameras each.
| Feature | ADT / SimpliSafe / Vivint | Cyrano HDMI tap |
|---|---|---|
| Monthly monitoring fee (single home) | $15 to $50 | $200 flat per property, any camera count up to 25 |
| Video add-on per camera per month | $10 to $25 | $0, included |
| Hardware required to start | Their panel, their sensors, their cameras | $450 device, keep your existing NVR and cameras |
| Install | Scheduled tech visit, hours | HDMI swap, under 2 minutes |
| Alert verification | Zone label + phone tree | On-device classifier + 480x270 JPEG of the tile |
| Where video lives | Their cloud, per-device storage plan | Your NVR's existing disk, thumbnails only to cloud |
| What happens during a Wi-Fi outage | Cloud cameras go dark; panel falls back to cellular | Detection continues on-device, queue drains on reconnect |
When the contract stack is still the right answer
I am not claiming the panel-and-central-station model is obsolete. There are homeowners for whom it is still the right fit, and it is worth naming them.
Where the panel stack still wins
- - You do not currently own cameras and you want one vendor to install everything.
- - Your insurance policy offers a material discount tied to a UL-listed central station.
- - You want glass-break, smoke, and carbon-monoxide detection in the same contract; the HDMI tap is camera-side only.
- - You live somewhere with armed-response ordinances that only honor the paid dispatch path.
Everything else, and particularly any homeowner who already has an NVR in a closet, is a case where the tap is the better fit.
Brands the template library already knows about
If your recorder is on this list, the install is the HDMI swap and nothing else. If it is not on this list, send in a photo of the multiview and we will add the overlay_mask template. The interface is the screen, so every recorder that draws a grid is covered eventually.
See your existing NVR become a monitored system.
Fifteen-minute call. Bring the brand of your recorder and a photo of your multiview. We will walk through exactly what the overlay_mask blanks for your layout.
Book a call →Frequently asked questions
What do the top search results mean by 'security systems home monitoring'?
They all mean the same bundle: a branded alarm panel (ADT Command, SimpliSafe Base Station, Ring Alarm Pro, Vivint SkyControl), a handful of door contacts and motion PIR sensors, optionally a couple of cloud cameras, plus a professional monitoring contract that routes panel trips to a UL-listed central station for $15 to $50 per month. The camera side is almost always a separate add-on with per-device cloud storage fees. None of the top results describe a path where the existing camera DVR or NVR becomes the monitoring source on its own.
I already have a DIY NVR system from Costco. Why is that not 'monitored'?
Because 'monitored' in the industry sense specifically means a human in a central station is paid to answer when the panel trips. A DIY Lorex, Amcrest, Reolink, Swann, or Night Owl NVR records 24/7 to its internal disk and shows a multiview on HDMI-out, but nobody is watching the feed and nothing automatically classifies what it sees. The footage is an archive, not an alert system. The gap between 'I own cameras' and 'my cameras are monitored' is the classifier. That is exactly what an HDMI edge-tap closes.
What is the HDMI edge-tap?
A small device that plugs into the HDMI output of your existing DVR or NVR and receives the same composite multiview the recorder draws for its wall monitor. Video passes through to the monitor unchanged, so the guard or homeowner still sees the same 2x2, 3x3, or 4x4 grid. The tap grabs frames from the passthrough and runs object detection plus four filter stages on the device itself. Classified events emit to a cloud inbox with a 480x270 JPEG of just the triggering tile. Nothing about the raw video leaves the house.
Which home NVR brands are actually supported by the overlay_mask templates?
Fourteen are covered out of the box, and every common consumer brand is in the list: Lorex, Amcrest, Reolink NVR, Swann, Night Owl, Q-See, ANNKE, EZVIZ, Hikvision DS-7xxx, Dahua XVR and NVR, Uniview, Bosch DIVAR, Honeywell Performance, and Panasonic WJ-NX. The overlay_mask blanks three things the recorder draws over the video before inference: the wall clock, the camera-name strip along the bottom or top of each tile, and the channel bug in the corner. If the recorder's multiview is readable on a monitor, it works.
What exactly lands in my inbox when an event fires at 2 AM?
A 9-field event JSON plus one image. The fields: tile.label (the camera name the NVR itself stamps on the strip, e.g. 'Back Porch' or 'Driveway'), tile.index (the row-major position of that tile in the grid), tile.coords (x, y, w, h in composite-frame pixels), property (which site or house this device belongs to), layout_id (the recorder layout, e.g. '2x2-std' for four cameras or '4x4-std' for sixteen), overlay_mask (the list of overlay classes blanked before inference), event_class (person_in_zone, vehicle_dwell, loiter, or tamper), iso8601_ts (the recorder clock timestamp), and latency_ms (capture-to-delivery time). The image is a 480x270 JPEG crop of just the triggering tile.
Why 0.058 percent? What does that number actually mean for my house?
It is the fraction of candidate detections that survive the filter pipeline and actually reach you. On a 25-tile multiview running continuously, the raw detector produces roughly 412,803 candidate bounding boxes per 24 hours (swaying branches, passing cars on the street, shadows, streaming-stick loading animations on an adjacent monitor). Four filter stages run before anything crosses the network: an overlay_mask that hides the recorder chrome, a tile-grid zone gate that only emits events inside zones you configured per tile, a multi-frame persistence check that requires the same object to appear for N consecutive frames, and a tile.label dedup that collapses repeats on the same camera. After all four, 241 events per 24 hours remain. That is what matters at 2 AM, because one real alarm in a pile of 411,000 drops is a product you can actually use.
Does this need the internet? What happens if my Wi-Fi drops?
Detection does not need the internet. Object classification, overlay masking, zone gating, persistence, and dedup all run on the device at your house. The thumbnail is generated locally. During an outage, events go to a local on-device queue and the JPEGs are written to a local buffer. When the link returns, the queue drains to the cloud in timestamp order and the inbox backfills. Contrast this with a cloud camera subscription (Ring, Nest, Arlo, SimpliCam): when the upstream fails, the camera effectively goes dark from your phone's perspective. Contrast it with a panel monitoring contract on cellular: when the cell link fails, the central station gets nothing.
Is this cheaper than ADT, SimpliSafe, or Vivint monitoring?
Over a typical home horizon, yes. A professional monitoring contract runs $15 to $50 per month plus video add-on fees of roughly $10 to $25 per camera per month. A four-camera home easily reaches $60 to $130 per month on the traditional stack. The HDMI edge-tap is $450 one-time for the device and $200 per month for the property, independent of camera count up to 25 on a single unit. The homeowner comparison that matters is that you do not need their panel, their installer, their sensors, or their cloud storage plan. Your existing DVR and your existing cameras are the system.
Will the police actually respond to a non-central-station alarm?
In most municipalities, yes, with one caveat. About 200 cities (Los Angeles, Dallas, Phoenix among them) charge $50 to $400 per unverified police dispatch on repeat offenders and a growing number (Salt Lake City, Seattle, and Las Vegas are precedents) require 'Verified Response' before dispatch. A pre-verified video alarm is precisely what those ordinances accept as verification: you call the non-emergency line with the event_class, the timestamp, and the thumbnail URL, and the dispatcher sees a real person in a real zone at a real timestamp before the patrol car is sent. A panel-only contract fails verification by definition because the central station only has a zone label.
Can I run this alongside an existing ADT or SimpliSafe contract?
Yes, they do not conflict. The panel side still monitors door contacts and motion PIRs and still reaches whichever central station is on your account. The HDMI edge-tap runs on the camera side in parallel. Many homeowners run both for a month or two, compare the alert volumes and the quality of thumbnails versus phone-tree zone labels, then drop the panel contract once they are comfortable. Nothing about the tap requires that you do anything to the panel side first.
Adjacent reading
If this angle fits, these will too
Security alarm monitoring when the alarm is already verified
What changes when the camera, not the panel, is the sensor and the alarm ships with a thumbnail already attached.
Home security camera system guide
How to choose a system you will actually use, and the install-and-forget trap most homeowners fall into.
Legacy CCTV real-time monitoring gap
The specific failure mode of a recorder that records twenty-four hours a day but alerts zero times.
Comments (••)
Leave a comment to see what others are saying.Public and anonymous. No signup.