Understanding Live Streaming for Surveillance
Live surveillance lives and dies by one number: the gap between something happening at the camera and you seeing it on screen. That gap is latency, and the smaller it is, the faster you can react. AI Live Insight is built to keep it as close to zero as your network allows, so the picture in front of you is the scene as it is right now, not as it was a few seconds ago.
This article explains how that low-latency video reaches you, why AI Live Insight uses two streaming protocols instead of one, and how you can scrub back through recent footage without ever leaving the live view.
Why Low Latency Matters
Think about what an operator does with a live feed. A weapon appears at an entrance, a vehicle pulls into a restricted lot, someone walks into frame after hours. Every second between that event and the video on screen is a second of lost reaction time.
Ordinary video streaming isn't built for that. To play back smoothly, a normal player buffers several seconds of video ahead of time, soaking up network hiccups so nothing stutters. That trade is perfect for a training video or a recorded webcast, where a few seconds of delay costs nothing. For a live camera you're watching to make a decision, those same seconds work against you.
Surveillance streaming makes the opposite trade. It keeps buffers small and prioritizes getting each frame to you as fast as possible, even on a busy network.
How AI Live Insight Delivers Live Video
There's no single streaming method that's both the fastest and the most reliable on every network, so AI Live Insight uses two and leads with the faster one. The player tries to connect over WebRTC first, and if that path isn't available, it falls back to Low-Latency HLS on its own. You don't choose between them in the moment; the player handles the handoff.
WebRTC for Sub-Second Latency
WebRTC is the same real-time technology behind live video calls, and it's the fastest way to move the picture from the camera to your screen. It opens a direct, continuous connection and pushes video through with almost no buffering, which gets latency down to under a second. That's why the player reaches for WebRTC first: when it connects, the feed is as close to real time as it gets.
Low-Latency HLS as the Backup
Some networks, firewalls, or proxies block the kind of connection WebRTC needs. When that happens, the player switches to Low-Latency HLS (LL-HLS) without interrupting your view. LL-HLS sends the video as a stream of very small pieces that arrive and play in quick succession, which keeps the delay to a few seconds, far less than ordinary streaming. It isn't quite real time, but it works almost everywhere, so you keep watching even where WebRTC can't reach.
The switch is automatic and silent. If WebRTC drops mid-session, the player moves to LL-HLS and the feed keeps running.
Object Detection on the Live Stream
The live feed isn't only streamed to your screen, it's analyzed as it arrives. While a camera is live, the surveillance service runs AI object detection on the stream and looks for the objects you chose, such as a person, vehicle, or weapon. Detection happens in step with the video, so what the AI flags lines up with what you're watching.
You see the results in two places. Turn on the Overlay in the player and detection boxes appear over the video, framing each object the moment it's recognized. On the timeline below the feed, the same detections show up as markers, so you can spot activity at a glance and jump straight to it. The detection data stays aligned with the video whether you're watching live or seeking back through recent footage.
Because detection rides along with the live stream, a match can do more than draw a box. It can raise an alert and trigger a recording, all from the same feed. To learn what the AI recognizes and how confidence and severity shape what gets flagged, see How Live AI Detection Works.
Choosing a Playback Mode
Most of the time you can leave the protocol decision to the player. When you need to, the player offers three modes that set which path it uses.
| Mode | What it does | When to use it |
|---|---|---|
| Auto | Tries WebRTC first, falls back to LL-HLS. | The default. Best for most cameras and networks. |
| Low Latency | Uses WebRTC only. | When you need the lowest possible delay and know WebRTC works on your network. |
| Standard | Uses LL-HLS only. | When WebRTC is blocked, or you want the most compatible, steadiest stream. |
Auto gives you the speed of WebRTC with LL-HLS as a safety net, which is why it fits nearly every situation. Reach for the other two only when you have a specific reason to pin the stream to one protocol.
Seeking Back Without Leaving Live
A live feed answers "what's happening now," but you'll often need "what just happened." AI Live Insight keeps recent footage ready behind the live edge, so you can drag back along the timeline, review the last few minutes, and return, all without opening a separate recording.
When you've seen what you need, select GO LIVE to snap back to the live edge and pick up the real-time feed again. This recent-footage buffer is separate from the longer-term recordings captured around alerts. To see how seeking, detection markers, and playback controls work in the player, and how recordings are stored for the long term, follow the related topics below.
What Affects Latency
The delay you see depends on more than AI Live Insight. The camera's own encoding settings, the network between the camera and the portal, and the distance the video travels all add fractions of a second. WebRTC gives the best case, but a congested or restricted network can still push the feed onto LL-HLS or stretch the delay. If a camera feels slower than expected, the network path is usually the first place to look.
Related Topics
- About AI Live Insight
- How Live AI Detection Works
- Understanding Live Camera Playback
- Understanding Surveillance Recordings