Glossary / Standards

What is ONVIF?

ONVIF is an open specification that lets IP video products from different manufacturers work together. It defines profiles, each covering a set of capabilities such as streaming, recording or analytics metadata. A device supporting a profile can be discovered and configured by any software supporting the same profile, without vendor specific integration work.

Reviewed by Dr. Vivek Gujar, Chief Strategy Officer  ·  Updated 31 August 2026  ·  6 min read

Before ONVIF, connecting a camera from one manufacturer to a recorder from another meant writing a driver. ONVIF replaced that with a common set of web service interfaces, so software can discover a device, ask what it can do, retrieve its stream URIs and configure it, all through one vocabulary.

The important thing to understand is that ONVIF is not a single capability. It is a set of profiles, and a device supports some and not others. A camera and a piece of software interoperate only where their supported profiles overlap.

The profiles that matter for video

ONVIF profiles and what each one covers

Profile	Covers	Why it matters
S	Video streaming and configuration, PTZ control, audio	The baseline. If a camera supports anything, it supports S. Enough for discovery and pulling RTSP.
G	Recording, storage and retrieval of recorded video	Lets software search and export from a recorder it did not build.
T	Advanced streaming: H.265, motion alarms, imaging settings, analytics metadata streaming	The modern replacement for S. Matters if you want H.265 or metadata alongside the video.
M	Metadata and events for analytics: object classification, geolocation, event subscription	Lets a camera's own analytics publish results in a standard form that other software can consume.
A, C, D	Access control configuration, door control, access peripherals	Relevant when video and access control share a system.

For a retrofit analytics project, Profile S is usually sufficient, because the job is to discover cameras and obtain stream URIs. Profile T is a bonus where the estate is newer. Profile M matters only if you intend to consume analytics the cameras themselves produce, which in a mixed estate is rarely the plan.

Conformant and compliant are different words

Check the list by model number, not by brand. Manufacturers register some models and not others, and a range that shares a marketing name can contain both. This takes two minutes and prevents a category of unpleasant discovery during commissioning.

What ONVIF does not guarantee

Even with genuine conformance, expect gaps. The specification defines the interface, not the completeness of a manufacturer's implementation, and several things fall outside it entirely:

• Feature depth. A camera may expose a setting through ONVIF that behaves differently from the same setting in its own web interface.
• Firmware variance. Conformance was tested on a specific firmware version. Older or newer firmware on the same model can behave differently.
• Vendor extensions. Manufacturer specific features live outside the standard and are unreachable through ONVIF alone.
• Image quality. ONVIF says nothing about how good the picture is, which is governed by the sensor, the lens and the encoder.
• Analytics quality. Profile M defines how results are published, not whether they are any good.

The practical stance: treat an ONVIF badge as evidence that integration will probably be straightforward, not as a contract. Discovery on the actual network, against the actual firmware, is the only real test. On an inherited estate this is worth doing before the project is scoped rather than after.

Discovery in practice

ONVIF discovery uses WS-Discovery, which is multicast and therefore does not cross subnet boundaries by default. On a flat camera VLAN a discovery scan finds everything in seconds. On a segmented network, cameras in other subnets have to be added by IP address, and the discovery step silently returns fewer devices than exist. A discovery scan that reports 40 of your 60 cameras usually means a routing boundary, not 20 broken cameras.

Frequently asked questions

Is ONVIF the same as RTSP?

No. RTSP is a streaming control protocol. ONVIF is a broader interoperability specification that covers device discovery, capability queries, configuration and events, and among other things tells software where to find a device's RTSP streams. You can pull RTSP from a camera with no ONVIF support at all, provided you know the URL.

Which ONVIF profile do I need for video analytics?

Profile S is normally sufficient for a retrofit, because it provides discovery, capability queries and stream URIs. Profile T adds H.265 and metadata streaming and is worth having on newer estates. Profile M matters only if you plan to consume analytics results the cameras generate themselves rather than running your own models.

What is the difference between ONVIF conformant and ONVIF compliant?

Conformant means the specific model passed ONVIF's test tool and is registered on the conformant products list. Compliant, supported and ready are marketing terms with no formal definition behind them. Verify by searching the conformant products list for the exact model number before relying on the claim.

Will ONVIF let any camera work with any recorder?

Only where their supported profiles overlap, and only for the features the profile covers. Basic streaming and recording generally work. Vendor specific features, advanced imaging controls and proprietary analytics usually do not. Expect the common denominator, and test before committing to a mixed estate design.

Do old cameras support ONVIF?

Many cameras from roughly 2012 onward support Profile S, though implementations from that era are often partial. Cameras older than that may support no version of ONVIF, in which case integration falls back to a known RTSP URL pattern for the model. This usually still works, it just cannot be discovered automatically.

Why does ONVIF discovery miss some of my cameras?

Discovery uses multicast, which does not cross subnet boundaries without explicit configuration. Cameras on a different VLAN from the scanning device will not appear, even though they are reachable by IP. Add them manually by address, or run the scan from inside each segment.