C
Cyrano Security
11 min read
Home Doorway Camera Guide

A home doorway surveillance camera is the wrong unit of thinking

Every Ring, Arlo, Nest, and Reolink page on this topic sells you one doorbell. That is fine for a porch pirate who walks straight up the front path. It misses the person who knocks, backs up two feet off the doorbell's motion cone, and waits. The setup that actually catches that visit is a small cluster of cameras around the doorway, all processed together by one edge AI device.

This guide walks through the multi-angle layout, the AI piece that makes it manageable, and how to run it on cameras you may already have without replacing any of them.

See multi-angle alerts in a live demo
4.9from 50+
Works with any existing DVR/NVR
Up to 25 camera feeds per device
2-minute HDMI install
$450 upfront, $200/month

Why doorbell cameras miss the visit you actually care about

Doorbell cameras have a single, narrow PIR motion cone pointed straight out from the door, typically 30 to 40 degrees wide at chest height. Inside that cone, they work well. Outside it, the camera has no idea someone is there. That is fine for a delivery driver who walks right up, but the loiterer who concerns most homeowners does not walk right up. They approach, knock, step back into the porch corner or the flowerbed to the side, and wait.

A single doorbell cam in that situation records the knock and the person leaving. It does not record the 40 seconds of waiting that is actually the useful signal. The fix is not a better doorbell. It is a cluster of cameras whose motion cones overlap around the entire approach.

One doorbell cam vs. a doorway cluster

One camera, one motion cone, one chest-height angle. The loiterer learns the boundary and stays outside it. Alerts fire for the knock and the departure. The wait in between is invisible.

  • Blind to anything 2+ feet off-axis
  • Captures the event, not the dwell
  • No correlation with other cameras
  • Cloud subscription per device

The five or six cameras a doorway actually needs

Nothing here is exotic. These are the angles that recur across every home-security walkthrough. What changes is how you wire them: into one recorder, with one AI device watching the whole cluster.

1. Porch camera (wide)

2.8mm or 4mm lens, mounted at head height 6-10 feet back from the door, angled down ~15°. Captures faces of anyone standing at the door. This replaces what the doorbell can't do well at close range.

2. Side-gate camera

Covers the fence-line path between front yard and back. Legitimate traffic here is rare, so sensitivity can be high and dwell thresholds short. This is the camera that catches the person who hops the gate.

3. Driveway approach

Points outward from the house toward the street. Catches vehicles pulling up and pedestrians walking past. Serves as the earliest-warning angle, often firing 10-20 seconds before the porch camera sees anything.

4. Package drop zone

Only needed if deliveries land somewhere the porch cam doesn't already cover, behind a pillar or screen door. A cheap wide-angle pointed at the drop spot runs the same dwell and object-removal detection.

5. Garage corner

Mounted at the garage-facing corner of the house, aimed along the driveway and toward the side path. Fills the blind spot between the driveway cam and the side-gate cam. Often the angle that gets the clearest walk-away shot.

6. Back deck or rear door (optional)

Most doorway incidents are front-of-house, but if your back door or deck is accessible from the street, a sixth camera here closes the last loop. Optional for most homes.

How the cluster feeds one AI device

Every camera in the cluster runs into your existing DVR or NVR. The DVR produces an HDMI multiview signal, the same one you would see on a wall monitor. An edge AI device reads that signal, splits it back into individual tiles, and runs detection on each tile. One box, six cameras, zero per-camera configuration.

Doorway cluster into one edge AI device

Porch cam
Side gate
Driveway
Package zone
Garage corner
Cyrano Edge AI
Loitering alert
Package removed
Dashboard

One device, six angles, one install

The specific piece that makes the multi-angle setup practical is that one Cyrano unit handles up to 25 feeds over a single HDMI cable from the DVR you already own.

0Camera feeds per device
0Cameras a typical home needs
0Install time (minutes)
$0Upfront hardware cost

The 25-feed capacity matters because it means you can over- provision your doorway cluster without ever worrying about hitting the limit. Six cameras is the realistic ceiling for a single-family home doorway. That leaves 19 feeds unused.

The one technical thing that makes this setup uncopyable

Most cloud smart-doorbell services process each camera in isolation. They cannot tell you "the person who was in the driveway 14 seconds ago is now on the porch." Edge AI at the DVR sees every feed simultaneously with sub-second latency, so it can stitch a single visit across cameras and fire exactly one alert for it.

Why correlation only works at the DVR/NVR level

The DVR is the one place in a home camera system where all feeds already converge. Every camera is already plugged into it. Adding AI at that point means the device gets the full multi-camera view for free, no extra wiring. Cloud services would need to re-aggregate feeds over the internet, which is expensive, high-latency, and why no cloud doorbell brand does true cross-camera tracking today.

Concretely: the porch cam fires at 7:42:11pm, 70% confidence on "person." The driveway cam had a 68% detection at 7:41:57 pm. The edge AI sees both within 14 seconds of each other and flags one loitering event that lasted ~45 seconds across two cameras. Operator gets a single notification with both clips attached.

See a multi-angle doorway setup in your own DVR

15-minute live demo. Bring your existing camera layout. We'll show what a cross-camera loitering alert looks like on your actual feeds.

Book a demo

What happens from the moment someone walks up

A cross-camera loiter event, as the edge AI sees it

01 / 05

t=0s — Driveway cam, first detection

A figure appears at the edge of the driveway. Confidence is low (63%) because the subject is at range. The system logs the detection but does not alert.

How to set it up in an afternoon

Assuming you already have cameras running to a DVR or NVR with an HDMI output, this is the actual sequence. If you do not, you will spend more time on the camera install than the AI piece.

1

Step 1 — Audit your current camera angles

Walk the property. For each approach path to your door, ask: which camera sees the first 10 feet, which sees the last 3? If the answer is 'none' for any path, that is a gap. Most homes discover 2-3 gaps they had not noticed.

2

Step 2 — Fill the gaps with cheap wired cameras

Add whatever cameras close the gaps. They do not need to be fancy. A basic wired camera that feeds into your existing DVR is fine. The AI pipeline does not care about per-camera features because detection happens at the recorder level.

3

Step 3 — Plug the edge AI device into the DVR's HDMI

Cyrano connects to the DVR's HDMI output port. The signal is passed through to your wall monitor unchanged, so you do not lose anything. Install is under 2 minutes. No DVR password, no IP credentials, no rewiring.

4

Step 4 — Configure dwell zones per camera

In the dashboard, mark the zones that matter on each camera tile. Porch: wide zone, 60-second dwell. Side gate: narrow zone at the gate itself, 30-second dwell. Driveway: full driveway, 90-second dwell. This takes about 10 minutes total.

5

Step 5 — Let the system run for a week, then tune

The first week will produce false positives specific to your house (a bush moving in the wind, the pool cleaner, the mail carrier). Dismiss them from the dashboard. The system learns from your feedback, and week-2 false positive rates typically drop by 40-60%.

How this compares to the usual doorway camera advice

FeatureSingle smart doorbellDoorway cluster + edge AI
Covers the whole approach, not just the doorstepNo, single motion coneYes, 4-6 overlapping angles
Catches the knock-and-dip visitor outside the coneNo, subject is invisible off-axisYes, side and driveway cams pick up
Uses cameras you already ownNo, you buy their hardwareYes, any DVR/NVR with HDMI
Sends one alert per loitering eventNo, each camera alerts separatelyYes, cross-camera correlation
All processing happens on your propertyNo, cloud analysis with upload latencyYes, edge AI, video never leaves
Flat monthly cost regardless of camera countNo, per-camera subscriptionYes, one fee for the device
Works if the WiFi is flakyDepends on WiFi strengthYes, wired DVR path

The three filters that keep alerts usable

A multi-angle setup without filtering produces alert fatigue in about two days. What makes it livable is the detection pipeline the edge AI runs on each feed.

Detection pipeline on every camera tile

1

Grid the scene

Divide the fixed camera view into regions with known background appearance and expected object sizes. One-time calibration per camera, done during install.

2

Require multi-frame persistence

A detection has to persist across 3-5 consecutive frames (about 1-3 seconds) before it counts. This alone cuts false positives from reflections, insects, and sensor noise by about half.

3

Apply dwell-time threshold

Per-zone dwell requirement. The person has to stay in the zone for the configured time (60s porch, 30s side gate) before an alert fires. Delivery drivers, mail carriers, and short-pause visitors never trigger it.

The camera choices that actually matter for a doorway

Specs worth caring about

  • Head-height mount (6-10 feet back from the door, angled down ~15°)
  • 2.8mm or 4mm wide-angle lens on the porch cam
  • Color night vision, not just infrared (clothing/hair color matters for ID)
  • Wired to the DVR (PoE or coax), not WiFi-dependent
  • At least 14 days of DVR retention, 30 is better
  • DVR with an HDMI output (any model from ~2012 forward)
  • Ability to export a clip as MP4/MOV in under a minute

What one deployment actually caught

In the first month at a property in Fort Worth, a single edge AI device added to the existing DVR caught 0 incidents across cameras that had been recording but never alerting before. One of those was a break-in attempt. The detection pattern was the same thing described above: a visitor who lingered past the dwell threshold was flagged, an operator reviewed the clip in under a minute, and escalation happened within the same window the visit was still in progress.

That deployment was a small multifamily property, not a single-family home, but the architecture is identical. One DVR, one HDMI cable, one AI device, a handful of cameras. Scaled down for a house, this is still the right shape.

Frequently asked questions

Why isn't one doorbell camera enough for a home doorway?

A video doorbell has a narrow motion cone pointed roughly straight ahead. The loiterers property owners worry about, people who knock, step back into the dark, and wait to see if anyone is home, quickly learn where the cone ends. Two feet off-axis and they are invisible to the doorbell. A multi-angle setup (porch cam, side-gate cam, driveway approach, package drop, garage corner) overlaps those cones so any approach path triggers at least one camera. One camera solves detection in theory. Three to six cameras solve it in practice.

Can I add AI loitering alerts to the doorway cameras I already own, or do I have to buy new ones?

You can add AI to the cameras you already own if they route through a DVR or NVR with an HDMI output, which covers almost every wired security system sold since roughly 2012. An edge AI device like Cyrano plugs into the recorder's HDMI port, reads the multiview signal, and runs detection on every camera tile at once. No camera replacement, no rewiring, no cloud subscription. One device handles up to 25 feeds, which is far more than any single-family home needs.

What dwell time should I use for doorway loitering alerts?

Start at 60 seconds for the porch and package zone, and 30 seconds for side paths and gates. 60 seconds filters out delivery drivers, mail carriers, and short-pause visitors. 30 seconds on side paths catches anyone who shouldn't be there at all, because legitimate traffic on a side gate is rare. Expect to re-tune after the first week based on the false-positive log. If you get alerts on the mailman every afternoon, push the porch threshold to 90 seconds. If you miss a real loiterer, drop the side-gate threshold to 15.

Will the camera capture someone's face if they are standing right at the door?

Only if it is mounted at roughly head height or slightly above, angled down about 15 degrees, and 6 to 10 feet away from the door. A camera mounted up under the eaves gets the top of everyone's head. Doorbell cameras are mounted at chest height, which gets faces beautifully for people standing 3 feet back, but distorts them at point-blank range. This is the single biggest reason the multi-angle setup works: a wide porch camera at head height does what the doorbell cannot, and vice versa.

How does this differ from cloud-based smart doorbell subscriptions like Ring Protect or Nest Aware?

Cloud subscriptions upload video to a remote server, run AI there, and stream alerts back. This adds 2 to 5 seconds of latency per frame, breaks multi-frame persistence filtering, limits cross-camera correlation (each camera is usually processed in isolation), and costs a recurring fee per camera. Edge AI on a DVR processes all feeds locally at sub-second latency, correlates events across cameras, and runs on a flat monthly fee regardless of how many cameras you have. For a doorway problem, the latency and correlation differences matter more than most people realize.

Is it legal to record someone standing at my front door?

In almost every US state, yes. A camera capturing activity in or near your own doorway, whether public-facing sidewalk, porch, driveway, or yard, is lawful. The exception is audio, which in eleven states requires all-party consent. The safest default is to disable audio recording and rely on video only. If the camera can see into a neighbor's window, reposition it. Doorway-facing cameras almost never have this problem.

Do I need a separate camera for packages?

Usually no, if your porch camera already covers where packages land. A separate package-zone camera is only worth it when deliveries drop on the side of the house, in a vestibule behind a screen door, or behind a pillar that blocks the main porch view. In those cases a cheap wide-angle camera pointed at the drop spot pairs with the rest of the system and gets the same AI dwell-time analysis. The point is that coverage is a design choice, not a product SKU.

The doorway is a cluster, not a camera

If you already own cameras, Cyrano plugs into your existing DVR in two minutes, covers up to 25 feeds on one device, and runs dwell-time loitering detection on every angle. $450 upfront, $200 per month, no per-camera fees, no cloud uploads.

Book a 15-minute demo
🛡️CyranoEdge AI Security for Apartments
© 2026 Cyrano. All rights reserved.

How did this page land for you?

React to reveal totals

Comments ()

Leave a comment to see what others are saying.

Public and anonymous. No signup.