Camera footage retention days are not a setting. They are an arithmetic result of three numbers you control — camera count, recording resolution, and how many hours a day each camera actually writes. Change any one and the retention window moves. Chasing a “90-day” label on a cloud plan tells you what the vendor will store, not what your own hardware can hold.
The formula below is the one I reach for whenever a buyer answers that they want thirty days and four cameras. Two minutes of work, and it has kept more people off the wrong drive than any specification sheet.
Days of retention = usable capacity ÷ (cameras × daily write per camera)
Everything else is a way of estimating the middle term. Two examples, worked in full, are further down.
Key Takeaways
- Continuous 4MP recording writes roughly 25 to 40 gigabytes per camera per day. A 4-camera system on thirty-day retention needs about 3 to 5 terabytes of usable space, not raw.
- RAID is not free storage. A four-bay array in RAID 5 gives you three drives’ worth of capacity; in mirrored pairs you get half. Budget the array, then compute retention.
- Motion-triggered recording can cut daily write volume by half or more versus continuous — but it also means missing the twenty seconds before the trigger that usually matter.
- A 256GB card in a camera is a buffer, not an archive. At continuous 4MP it holds a handful of days at best.
The Short Answer: How Do I Work Out Days of Footage?
Take usable capacity in terabytes, divide by the total gigabytes your cameras write each day, and divide that result by 1000 to convert gigabytes into terabytes correctly. If you would rather skip the unit juggling: one terabyte holds roughly two hundred and fifty camera-days at 4MP continuous.
That last figure is the one worth memorising. One terabyte therefore holds around two hundred and fifty camera-days at the pessimistic end of the 4MP range, and up to four hundred at the optimistic end. Use 250 when you are sizing a purchase — underestimating retention is the error that costs you footage. A single 8-terabyte drive therefore holds at least two thousand camera-days, so four cameras on that drive gives you five hundred days of continuous footage at worst.
That number tells you something immediately: a single large drive is usually more retention than a home or small shop needs, and a two-bay box with mirrored 8-terabyte drives is a more sensible purchase than a four-bay box with small drives. For the common case, four 4MP cameras at thirty days land at roughly 4 terabytes usable, which two mirrored 4-terabyte drives cover.
Write the three inputs down before you shop. Camera count. Resolution. Continuous or motion-triggered. Everything on a product page is downstream of those — including how many bays you need and how large each drive should be.
Where Does the “Camera-Days per Terabyte” Number Come From?
From bitrate, and bitrate is set by resolution and compression. The figures below are engineering estimates for H.265 encoding at typical consumer settings, not measurements we ran in a lab. Treat them as planning values and expect plus or minus a third either way, because scene complexity and frame rate move them.
| Resolution | Continuous write per camera per day | Camera-days per terabyte | Reality check |
|---|---|---|---|
| 2K / 4MP class | roughly 12 to 20 GB | about 500 to 800 | The lowest write volume you can buy at this resolution |
| 4MP / 2K+ | roughly 25 to 40 GB | about 250 to 400 | The common default for driveway and shop floor |
| 4K / 8MP | roughly 60 to 100 GB | about 100 to 160 | Retention collapses; only use where identification matters |
The multiplier that surprises people is the jump from 4MP to 4K. Pixels quadruple, and even with better compression the daily write roughly doubles or triples. If the goal is a long retention window rather than facial detail at thirty metres, dropping the resolution buys you more than any hardware upgrade.
Is the Capacity on the Box the Capacity You Get?
No, and this is where most retention estimates go wrong by a third or more. Three deductions come off the raw total before a single frame is written.
- RAID overhead. A mirrored pair wastes half the raw capacity but survives a drive failure. RAID 5 across four drives wastes one drive. Parity is a deliberate trade: you are buying uptime with retention days.
- Filesystem and reserved space. Metadata, snapshots and the space a modern filesystem keeps in reserve for its own integrity checks all take a slice. Plan on losing a few percent before you start.
- The buffer you keep. Filling an array to the last byte is how you turn a full disk into a corrupt recording. Keep a working margin.
Work the numbers both ways. A four-bay enclosure holding four 8-terabyte drives has 32 terabytes raw. Configured as mirrored pairs you have 16 terabytes usable. Configured as RAID 5 you have about 24. That difference is not academic — it is roughly two hundred extra camera-days at 4MP, or the difference between a month of 4K footage and two months of 4MP footage.
If you want a hand with the array sizing, the woCyber NAS range is laid out by bay count so you can match the enclosure to the retention target rather than guessing.
What Does the Arithmetic Look Like for a Normal House?
Four 4MP cameras, continuous recording, thirty days of retention.
Daily write: 4 cameras × 30 gigabytes = 120 gigabytes per day. Thirty days needs 3.6 terabytes. Add a working margin and round up, and you are asking for about 4 terabytes usable. In a mirrored two-bay configuration that means two 4-terabyte drives; if you want headroom for a fifth camera later, two 8-terabyte drives cost less per terabyte and give you well over thirty days.
Now the same house on motion-triggered recording. A driveway camera that actually sees traffic twelve hours out of twenty-four writes roughly half the volume. The same 4-terabyte usable target now stretches toward fifty or sixty days, or lets you keep all four cameras at 4K.
The trade is the one nobody advertises. Motion detection decides what the recording is about. A person walking up a path triggers the clip — but the two seconds where they arrived on the edge of frame may not have triggered anything. Continuous is boring to review and complete. That completeness is what you are buying retention for in the first place.
What Does Each Retention Target Cost in Drives?
Here is the same arithmetic pre-computed for the usual targets, at 4MP continuous and using the conservative 250 camera-days per terabyte. Find your column, read the drive pair.
| Retention target | 2 cameras | 4 cameras | 8 cameras |
|---|---|---|---|
| 30 days | about 2 TB usable | about 4 TB usable | about 8 TB usable |
| 90 days | about 6 TB usable | about 12 TB usable | about 24 TB usable |
| 365 days | about 24 TB usable | about 48 TB usable | about 96 TB usable |
Read the last row and the retention-versus-resolution trade becomes obvious. Eight cameras held for a year at 4MP is an enterprise array. The same eight held for a year at 2K-class bitrates is roughly half that. Nobody needs to be told which is the better buy for a domestic site — the table tells them.
Why Do Cloud Plans Quote 30 or 60 Days and Local Storage Feels Open-Ended?
Because a subscription is priced per day of storage and a drive is priced once. That is a business difference, not a technical one. The vendor’s cost scales with how long you keep footage, so the plan caps it. A drive’s cost is sunk the day you buy it, so the retention limit is physical rather than commercial.
This matters for how you compare the two. A cloud plan at a fixed monthly fee for a fixed window looks like good value in month one and poor value in year three, because the fee does not stop. A drive purchase front-loads the cost and then runs. The honest way to compare is total cost across the period you actually intend to keep the system — and then ask what happens to your footage if you stop paying.
Reolink’s Home Hub 2 went on sale on 15 October 2026 with an advertised price of USD 199.99, marketing local on-device video search and no subscription. Read the vendor’s own product page for the current specification and price — those change. The strategic point is narrower: local retention is now a product category rather than a niche argument.
How Long Can You Actually Keep Footage on a Camera SD Card?
Days, not months. A 256GB card recording 4MP continuously receives roughly 30 gigabytes a day, so the arithmetic runs out somewhere between a week and ten days — and the card is then overwritten from the beginning.
The larger problem is endurance. Consumer microSD cards are rated for a limited number of write cycles and a surveillance workload spends them continuously. Cards in 24/7 cameras fail early, and they fail quietly. Two things follow. First, use high-endurance cards rated for surveillance if you must record on-card. Second, treat the card as a buffer for the last few days, with the NAS as the archive.
There is a structural reason the card matters more than it looks. A thief who takes the camera takes the card. A camera writing to a NAS elsewhere in the building does not lose its archive when the hardware walks away.
What About the Cameras That Never Reached a Network?
They still count toward retention, and they are the ones buyers forget. A solar camera on a fence line with a 4G link may buffer locally and upload on a schedule, so its effective retention is a function of both its card size and how often the link is available. A wired camera on your own network writes as fast as the drive accepts it.
Plan the two groups separately. Mains-powered cameras on a network can be sized with the formula above. Off-grid cameras need their own calculation, because the constraint is power and connectivity rather than drive space. The woCyber line builds both sides of that split — solar and 4G cameras for sites without mains power, and NAS storage for the archive — and the sizing conversation is easier when you bring both counts.
Solar and 4G cameras need a different calculation again.
One number that bites here: an off-grid camera that uploads in bursts needs a NAS sized for the burst, not the average. Four cameras delivered over a slow link will queue, and the queue needs a landing zone. In practice the units woCyber builds for that job are matched pairs — the NAS Security Kit pairs cameras with local recording so the archive starts on day one rather than after the first failed sync.
Where Do People Get the Retention Calculation Wrong?
Three errors account for most of the disappointment.
- Planning on raw capacity. A 4-drive 32-terabyte box in RAID 5 gives about 24 usable. Anyone who budgeted thirty days against 32 and configured parity will be short.
- Forgetting the resolution they will actually run. Sizing at 4MP and then switching a camera to 4K for the number plate doubles or triples its daily write overnight.
- Counting the retention window from the wrong end. The oldest footage is overwritten first. If an incident is discovered eleven weeks later and the window is thirty days, the recording is already gone. Decide the window from how long it takes you to notice a problem, not from what feels generous.
The third one is the expensive one. Retention is not about storage convenience; it is about the interval between something happening and you finding out.
Our Take
Buy retention, not capacity. Those are different purchases, and conflating them is why people end up with a full array holding four days of 4K.
The counter-intuitive judgement: for most homes and small shops, two mirrored drives beat four drives in RAID 5. You give up usable capacity, and in exchange you get the thing that actually protects footage — survival of a drive failure without a rebuild window. A four-bay enclosure only earns its cost when you genuinely need the retention days or the write throughput. Buying bays because they look more serious is how a retention plan turns into a hardware collection.
And the one place I would push back hardest: if you are buying a camera specifically to keep long evidence, spending the marginal budget on a bigger drive instead of a higher resolution will usually serve you better. 4MP held for ninety days outperforms 4K held for twelve, unless your job is identifying faces.
What We Don’t Know
The bitrate ranges above are planning estimates, not measurements taken on your site, and scene complexity can move them materially. A camera pointed at a busy road writes more than one pointed at a wall, at identical settings. Our ranges are deliberately wide for that reason.
They also assume H.265. Cameras still shipping H.264 write substantially more per day for the same picture — a correction factor of roughly 1.5 to 2 on the daily figures. If your camera does not offer H.265, treat every estimate in this article as the optimistic end and re-run the arithmetic before you buy drives.
Questions Buyers Ask Before Sizing Storage
How many days of footage can 1TB hold for a security camera?
At 4MP continuous, roughly 250 to 400 camera-days — so a single camera fills it in about eight to thirteen months, and four cameras in about two to three months. At 4K the same terabyte holds roughly 100 to 160 camera-days.
Is 30 days of camera footage storage enough?
Thirty days is the common default because that is what most cloud plans sell. Whether it is enough depends on your discovery lag: if nobody reviews footage until a dispute arises six weeks later, thirty days is already gone. Ninety days is the more defensible window if the system is meant for evidence.
Does motion detection really extend camera footage retention days?
Yes, often substantially, because the volume written scales with recorded seconds rather than wall-clock time. A camera seeing activity twelve hours a day writes roughly half of what continuous recording writes. The cost is completeness at the edges of an event.
Do I need RAID for camera footage retention?
Not for retention — RAID costs you capacity. You want it for availability, so a failed drive does not stop recording while you source a replacement. If the footage matters and the camera system runs unattended, mirroring is the cheaper risk reduction than redundancy across many drives.
How long do security cameras keep footage without a subscription?
As long as your own storage allows, up to the physical capacity of the drives installed. Without a cloud plan there is no vendor-imposed cap; the limit is the arithmetic in the formula above, plus the fact that the oldest recordings are overwritten first.
Do the arithmetic before you shop and the array sizes itself. Send us three figures — how many cameras, what resolution, how long you must keep the recordings — and we will come back with the bay count and drive capacity that covers it.
Related solution
woCyber NAS Storage
Bays, drives and RAID levels chosen for your camera count and retention window — sized before you buy, not after.
Not Sure How Much Storage You Actually Need?
Send us your camera count, resolution and how long you need to keep footage — we will size it for you.
- Storage sizing for IP cameras
- Retention planning (30 / 90 / 365 days)
- RAID and backup configuration
- Business & commercial applications
A typical cloud plan keeps billing: monthly fees stack up for as long as you keep the cameras.
woCyber records on your own hardware with zero monthly fees — one purchase, no plan, no renewal, no price hikes.
✓ What we will never do: no subscription lock-in, no footage uploaded to our servers, no selling your data — your recordings stay on hardware you own.