The short version
- DVR (Digital Video Recorder) — records analogue cameras. The camera sends a raw video signal over coaxial cable; the DVR does the digitising and encoding. The intelligence, such as it is, lives in the box.
- NVR (Network Video Recorder) — records IP cameras. The camera encodes video itself and sends it over the network; the NVR stores and manages the streams. Cameras and recorder are separate devices on a LAN.
- VMS (Video Management Software) — software rather than an appliance. It manages many cameras, and often many recorders and sites, with users, permissions, search and export. A VMS may run on ordinary server hardware.
The category boundary that actually matters is where encoding happens (DVR vs NVR) and whether you are buying a box or software (NVR vs VMS).
DVR — analogue in, digital out
DVRs persist because analogue coax is already installed in an enormous number of buildings, and pulling new cable is expensive. Modern HD-over-coax standards carry 1080p and beyond, so "analogue" no longer means "low resolution."
The real constraints are architectural: one cable per camera back to the recorder, a channel count fixed by the hardware you bought, and the recorder as a single point of failure for every camera attached to it.
NVR — IP cameras, network storage
With IP cameras the camera encodes H.264 or H.265 itself and streams over the network, typically powered by the same cable via PoE. That decouples camera count from recorder ports and lets cameras sit anywhere the network reaches.
The trade is that you now own a network problem. Bandwidth, switch capacity, VLAN design and multicast behaviour all become part of the video system, and a network fault is a video fault. It is also worth knowing that most NVR "analytics" are still motion-based — a pixel-change trigger with a friendlier name.
VMS — software that manages the whole estate
A VMS is what you reach for when the problem stops being "record these cameras" and becomes "operate video across sites, with roles, audit trails, retention policy and investigation workflows." It usually supports mixed hardware, multiple recorders and federated sites.
The cost is complexity and licensing — typically per camera, per site, or per feature tier — and a server to run it on. For a single building with eight cameras, a VMS is usually more machinery than the problem deserves.
Where edge AI is a different decision entirely
Here is the part the acronyms obscure: recording and understanding are separate concerns. A DVR, an NVR and a VMS are all fundamentally storage-and-retrieval systems. They answer "what was recorded at 3am?" They do not answer "was anyone there, and should someone have been told?"
Perception at the edge answers the second question, on the device, as the video arrives. That is why it is not a fourth box competing with the other three — it is a different layer, and it can sit alongside any of them:
- Alongside an existing NVR or DVR — the recorder keeps doing what it does; an edge perception node subscribes to the same camera streams and adds presence and dwell awareness with evidence.
- Instead of a small NVR — for a handful of cameras, an edge appliance that both records and understands removes a box from the design.
- Feeding a VMS — perception generates the events; the VMS remains the system of record for a larger estate.
Choosing between them
- Existing coax you cannot replace — a DVR, or an encoder that puts those cameras onto the network.
- New install, one site, modest camera count — IP cameras plus an NVR, or an edge appliance if you want understanding as well as recording.
- Many sites, many users, formal investigation workflows — a VMS.
- You need to be told when something happens — that is a perception requirement, and none of the three categories delivers it by default. Evaluate it separately and on evidence, not on a datasheet bullet that says "smart detection."
The question to ask a vendor
Ask what triggers an alert. If the honest answer is "pixels changed in this region," you have a motion detector, whatever the marketing says — and you will discover its false-alarm rate the first windy night. Ask what happens when a person stops moving, what happens to footage when the disk fills, and whether anything leaves the building.
Where ManasaView fits
ManasaView is an edge perception appliance rather than a recorder category. It records continuously, confirms presence rather than motion, attaches evidence to every alert, and keeps all of it on the device. It can run beside an existing NVR or replace a small one.
Frequently asked questions
What is the difference between a DVR and an NVR?
A DVR records analogue cameras and does the video encoding itself, usually over coaxial cable. An NVR records IP cameras that encode video themselves and send it over the network. The practical difference is where encoding happens and how cameras are cabled.
Is a VMS just a software NVR?
Not quite. A VMS manages video across many cameras, recorders and often multiple sites, adding users, permissions, search, retention policy and investigation workflows. An NVR is typically a single appliance recording a fixed set of cameras.
Do NVRs include AI or analytics?
Many advertise analytics, but a large proportion is still motion detection — a pixel-change trigger. That is why false alarms from wind, rain, shadows and headlights remain common. Ask specifically what condition raises an alert.
Can I add edge AI to my existing NVR or DVR?
Usually yes. Recording and understanding are separate concerns, so an edge perception node can consume the same camera streams while the existing recorder continues unchanged. That avoids replacing working infrastructure.
Do I need an NVR if I have an edge AI appliance?
Not necessarily. For a small site, an edge appliance that both records continuously and detects presence can remove a box from the design. For larger estates with formal investigation workflows, a VMS or NVR usually remains the system of record.
← Back to Knowledge