AI Vision MonitorHome
← Use Cases

Why Your Camera Misses Events

You saw the van pull up. You open the app an hour later and there is nothing in the timeline between 11:04 and 13:20. The camera was powered, online and pointed at the driveway the whole time. This is the most common complaint about home cameras and almost none of it is a broken camera. It is a chain of small delays and one sensor that was never designed to notice the thing you care about.

Two different failures that look identical in the app

An empty timeline means one of two things, and they have nothing to do with each other. Either the camera never woke up, so no frame of that moment exists anywhere. Or the camera recorded it perfectly and never told you, and the clip is sitting in a list you scrolled past. The fix for the first is hardware. The fix for the second is not buying anything.

Worth separating before you spend money. Open the storage rather than the notifications and look for the moment you know happened. If the clip is there, your problem is alerting. If there is no clip at all, your problem is the trigger, and a better recorder behind the same trigger changes nothing.

What happens in the seconds before a battery camera records

A battery camera is asleep almost all of the time. That is the whole reason it lasts months on a charge. Waking it up runs in order: the passive infrared sensor notices a warm object crossing its zones, the processor comes up, the radio associates with your wifi, and only then does a frame get written. On most battery models that chain takes somewhere between one and five seconds.

Everything that fits inside that window did not happen, as far as the camera is concerned. A courier who jogs from the van to the porch and back is inside it. So is a cat. So is most of what a person does on a doorstep. When people say a camera starts recording too late, this is the mechanism, and it is why the clip so often opens on an empty frame with a car door closing somewhere off screen.

Infrared sensors also have a preference that catches people out. They read change across their zones, so a person walking past the lens sideways is easy and a person walking straight at the lens is much harder. A delivery driver coming up a path directly toward the camera is the hardest case there is, which is exactly the case everyone buys the camera for.

The parcel problem, in one paragraph

A cardboard box is at room temperature and it does not move. To a heat sensor it is not an object, it is part of the wall. So the trigger fires on the courier, runs out of motion three seconds later, and nothing about the box itself will ever wake the camera again. The parcel can sit there for six hours and then leave in somebody's hands, and if that somebody was quick enough, the whole story is missing from the timeline. The camera is working exactly as built.

What you missedWhy the trigger did not fireWhat actually catches it
Parcel arriving while you were outCourier was inside the wake window, box has no heatA check of the porch on a schedule, asking whether a parcel is there
Parcel taken hours laterNothing warm moved across the zones near the boxThe same check noticing the porch is empty again
Someone walking straight up the pathHead on approach gives the sensor the least changeCamera angled across the path rather than down it
Car in the driveway at 3amToo far, too cool, or beyond the sensor range in the darkA check asking whether a car is parked there
Gate left open all afternoonOpening was one event and the state that followed is not motionA check asking about the state of the gate, not the movement of it
Event is in storage but you never saw the alertNothing failed, the notification lost a fight with forty othersFewer alerts, each one about a described condition

What replacing the hardware fixes and what follows you

The usual next step is to rip it all out and buy a wired system with an object detection layer on top. It is a real fix for a real part of the problem, so it is worth being precise about which part.

Ask about a state instead of an event

Every trigger based camera has the same shape of failure, because a trigger has to fire at the right instant or the moment is lost for good. A scheduled check has no instant to miss. It photographs the porch every ten minutes and answers the same question each time, which is not what moved, it is what is there now.

AI Vision Monitor is built on that second question. It takes still photographs on a schedule from a phone or a camera you already have, describes them against a sentence you wrote, and acts when the description matches. A parcel that arrived during a wake window still shows up on the next check, because the parcel is still on the porch and the check is not looking for movement.

Check the porch every 10 minutes. Tell me when a parcel or box appears, and tell me again if it is gone before I am home.
Alert me if the side gate is open. I do not care who opened it.
Check the driveway at 07:00 and 22:00 and tell me whether the car is there.

The pattern is the same in all three. You name a condition in ordinary words and you say when to look. Nothing in them depends on a sensor firing at the correct second, which is the part that keeps failing.

What that costs to run

Checks are paid one at a time rather than by the month. The starter pack is 2000 credits for 2.99 euro, a check on the light model costs half a credit and on the fast model one credit, and new accounts get 50 free credits without a card.

ScheduleChecks per dayCredits per day (light)A month
Every 10 minutes, 08:00 to 20:007236about 1.60 euro
Every 30 minutes, 08:00 to 20:002412about 54 cents
Every 10 minutes, delivery window only189about 40 cents
Twice a day, morning and night21about 4 cents

Most people who set this up for parcels run the third row, because the interesting hours are the ones they are out of the house and the rest of the day answers itself when they walk past the door.

Honest limits

Frequently asked

Why does my camera start recording late? Because a battery camera has to wake, boot and join the network after the sensor fires, which takes a few seconds. Anything that finishes inside that window never gets a frame.

Why did it miss the delivery driver but catch the neighbour's cat? Direction. Infrared sensors read movement across their zones, so something crossing the view is easy and something walking straight at the lens is close to invisible to them.

Why does nothing trigger on the parcel itself? A box is at room temperature and stationary, so there is nothing for a heat and motion sensor to detect. Only the person carrying it ever triggers anything.

Will a wired system fix it? A continuous stream removes the wake delay, which is the biggest single cause. It does not change what happens when a parcel sits still for six hours, and it does not reduce the number of alerts you ignore.

Will detection software fix my battery cameras? No. It can only work on what the camera streams, so the missing seconds stay missing. It does make the clips you do get much easier to search.

How is a scheduled check different? It never has to fire at the right moment. It asks the same question every ten minutes and answers from what is in the picture, so a thing that is still there gets reported whether or not anything ever triggered.

Related checks people run on the same phone

The missed delivery is usually what sends people looking, and the page on knowing when a parcel actually arrived works through that one case in detail. The opposite complaint, a camera that fires all day on rain and headlights, is covered in why motion alerts go off for nothing. If the bill is the question, the cost per day arithmetic has its own page.

Try it against the thing you keep missing

Take the event your current camera missed last week, write it as a sentence about a condition rather than a movement, and point a spare phone at the place it happens. Give it ten minute checks across the hours you are out. It costs a few cents to find out whether the question was ever the camera.

Get it on Google Play Download on the App Store Open in your browser