Media Router is the distribution half of the platform. It takes streams in over any common protocol and codec, publishes them back out in whatever the consumer requires, and sits as a boundary between your camera network and everyone who needs to watch, so each partner, portal, and mobile user gets exactly the feeds they are entitled to and nothing else.
Deployments run to 20,000 cameras and more, with health monitoring, alerting, and reporting across the estate.
Request Demo Read the Technical Brief
A proxy layer for camera video, built so that giving somebody access to a feed never means giving them access to the network the camera lives on.
Every agency that owns cameras eventually gets asked to share them. A neighboring department wants a few intersections during an incident. A news station wants the highway feeds. The public wants a portal. The path of least resistance is to hand out RTSP URLs and camera credentials, and that path ends with credentials in a spreadsheet, no idea who is watching, and a compromised partner becoming your incident.
Media Router replaces that with a controlled boundary. Streams arrive from Forge or directly from cameras, and go out on isolated endpoints for each consumer, authenticated per session with one time tokens or by IP where the consumer is a fixed system. Access is granted, changed, and revoked centrally, and every session is logged.
It is also a format engine in its own right. Every common video codec goes in and comes out again, H.264 through H.265, AV1, VP8 and VP9, MPEG-4, MJPEG and MPEG-1 or 2, with AAC, Opus, MP3, AC-3 and G.711 on the audio side, published over RTMP, RTSP, HLS, MPEG-DASH, SRT, WebRTC, and MPEG-TS. Where a receiving system will only talk to a camera, the router can present itself as one.
Around distribution sits the operational layer that a camera estate of this size needs: a searchable camera directory, tours, continuous health monitoring, alerting with configurable thresholds, live multi-camera viewing, coverage maps, and reporting that can be handed to somebody who does not work in video. For fleets that need formal governance on top, including inventory keyed to your VMS and independent verification of what partners can actually see, WINK Crossroad adds that layer from the cloud.
The default answer for web viewers is a one time password token, created per viewer session and appended to the playlist request. Nothing permanent is issued, so nothing permanent has to be chased when access ends.
Tokens are created through the API with a duration in minutes, extended while a session is still watching, and destroyed on demand. They are valid across every router in the deployment, so a viewer needs one token rather than one per node. Full detail is in the API Reference Manual.
For fixed systems such as a partner VMS or an emergency operations center, network level access is simpler than token handling. Whitelisted sources are still subject to behavioral monitoring, with automatic temporary blocking when usage patterns go abnormal.
A five tier permission model with per camera grants down to the individual user, session management, and partner organization sharing that does not require the partner to hold an account on your VMS.
Publish and playback are controlled separately. A publish ACL governs which addresses may push a stream into the router, and a playback ACL governs which may pull it out, both by IP or range.
Every stream carries a globally unique identifier with collision avoidance, so a stream name is never ambiguous across routes or routers. Transport is encrypted, permissions are checked per request, and access and configuration changes are both in the audit trail.
Two views: how the proxy layer isolates each consumer, and how one viewer session is authenticated across a multi router deployment.
| Input |
|---|
| MPEG transport stream over UDP and RTP |
| RTMP ingest |
| RTSP sources |
| HLS input |
| WebRTC sources |
| Output |
|---|
| MPEG-TS with RTP and UDP |
| HLS with dynamic segmentation |
| RTMP distribution |
| WebRTC output |
| MPEG-DASH delivery |
HLS with OTP authentication. Works in any browser, survives the public internet, and expires by itself.
RTSP with IP whitelisting. Native support in the receiving system, sub-second latency, no credentials in URLs.
SRT with a pre-shared key, where the receiving end has a compatible decoder and the quality bar is a production one.
A distribution layer that only speaks one codec pushes the conversion problem back onto the camera estate. Media Router accepts and emits the formats that are actually in service.
| Video | |
|---|---|
| H.264 / AVC | All profiles, baseline required for WHIP |
| H.265 / HEVC | Main and Main10 |
| AV1 | Next generation compression |
| VP9 | WebRTC and YouTube compatible |
| VP8 | Legacy WebRTC |
| MPEG-4 | Legacy cameras |
| MJPEG | Low latency, high bandwidth |
| MPEG-1 and MPEG-2 | Broadcast compatibility |
| Audio | |
|---|---|
| AAC | High quality, widely supported |
| Opus | WebRTC |
| MP3 | Legacy compatibility |
| AC-3 | Broadcast and theatrical |
| G.711, PCMU and PCMA | Telephony and intercom sources |
Input and output, so a deployment can standardize its public delivery without re-specifying its cameras.
Some VMS platforms will only accept a camera. Media Router can be that camera, responding as the manufacturer's own device would:
Emulation requires the appropriate licensing. Configuration detail is in the WINK Video Bridging Manual.
Where the estate needs an authoritative inventory keyed to your VMS, frame level health verification, and independent checking of what each partner can actually see, WINK Crossroad adds that governance layer on top of the routers.
Both add-ons process video directly from Media Router, so adding intelligence does not mean adding infrastructure or pulling a second copy of every stream.
Direct camera access is always the first thing a partner asks for. It is simple to explain and costs them nothing to implement. It is also the arrangement that produces credentials in a spreadsheet, connection counts nobody controls, and a security boundary that now includes an organization you cannot audit.
The property that matters is blast radius. When ten partners share one endpoint and one credential, a rotation affects all ten, so in practice it never happens. When each partner has its own endpoint, port, credential pair, IP whitelist, and camera set, ending one relationship is a single operation with no side effects. Per session OTP tokens extend the same idea to individual viewers: nothing long lived is issued, so nothing long lived can leak.
Web viewers get HLS with a token and accept a few seconds of latency in exchange for playing everywhere. A partner VMS gets RTSP with IP whitelisting, because that is native to the receiving system. Broadcast gets SRT with a key. One warning worth stating plainly, because it accounts for a large share of support cases: putting a general purpose CDN or web application firewall in front of a live video endpoint frequently breaks it.
The station needs twelve highway cameras for the duration of a closure. They get their own endpoint, their own credentials, and exactly those twelve cameras, with viewers authenticated per session. Their playout system will only accept an Axis camera, so the router presents itself as one. When the closure ends, access is withdrawn in one operation. No other partner notices, no camera credential changed, and the access log shows precisely what was watched and when.
More detail: the Camera Sharing and Partner Integration Guide, the API Reference Manual for token mechanics, and the Firewall and Network Configuration Guide for the port tables.
Tell us who needs access to what, and we will show you how the boundary is drawn.
Request Demo