What latency means in plain words
Latency is the gap between something happening on the stage and the same moment appearing on a viewer's screen. If a drummer hits the snare and a person at home sees it eight seconds later, the latency of that stream is eight seconds. It is not a fault in the usual sense; it is the sum of every step the picture has to pass through on its way out of the building.
Organisers often ask us for a stream with no delay. In practice nothing sent over the internet arrives instantly. The question worth asking is how much delay suits the event, because shortening it always costs something elsewhere, usually in stability or in picture quality.
We explain this early in planning, since the word gets used loosely. Some people mean the lag on the stream, some mean the delay between a cue and a graphic, and some mean the pause before a video starts playing. Each has a different cause and a different remedy.
Where the seconds go
The delay is built from several small pieces. The cameras and the vision mixer add a few frames. The encoder then compresses the programme feed into a form that can travel, and it needs a short run of video in hand before it can do that well. After that the stream is uploaded, split into small chunks by the delivery service, copied across servers, and finally downloaded by each viewer's player.
The player is the part people forget. To avoid stuttering, it deliberately holds back a few chunks before showing anything, so that a brief dip in someone's connection does not stop the picture. That protective cushion is what viewers call buffering when it runs out.
Added together, these steps explain why a stream that looks perfect on our monitors in the control room still arrives noticeably after the live moment.
- Capture and mixing: a few frames
- Encoding: a second or two
- Upload and delivery: several seconds
- Player buffer: the largest single share
Why shorter is not always better
It is possible to trim the delay by using smaller chunks, a smaller player buffer and a faster encoder setting. The trade is that the stream becomes less forgiving. A viewer on a weak mobile signal will see the picture freeze more often, because there is less spare video waiting in reserve.
For a concert or a stage production watched passively, we usually advise accepting a longer delay in return for a steady picture. Nobody at home is comparing the stream against the room, and a clean, unbroken image matters far more than being a few seconds closer to live.
A short delay earns its cost when the audience is interacting, for example answering questions sent in during a talk, or following a result that people are also hearing about elsewhere.
Keeping sound and picture together
A related term is lip sync. Audio and video are processed separately, and each can pick up a different delay on its way through the chain. If they drift apart, a singer's mouth no longer matches the voice, which viewers notice far sooner than a general lag.
We check this at the start of every stream by watching a known reference, such as a clapper or a clear on-stage action, against the encoded output rather than trusting the settings alone. If the sound arrives early, we hold it back by a measured number of milliseconds until the two agree.
This is also why we monitor the outgoing stream on a separate device throughout the show, not only the programme feed in the room. What the viewer receives is the thing that has to be right.
What to tell us when planning a stream
A few answers early on let us choose the right balance. Tell us whether viewers will be interacting with the event or simply watching it, because that decides how much delay is acceptable. Tell us roughly how many people you expect and what sort of connections they are likely to use, since a mostly mobile audience needs more headroom than one watching on office broadband.
Tell us also about the venue's own internet. Upload capacity at the site sets the ceiling for picture quality, and we would rather learn about a thin connection during planning than discover it during the soundcheck.
With these details we can set the encoder, the delivery settings and the player buffer to suit your event, and say plainly what delay to expect before the doors open.
- Interactive or watch-only audience
- Expected viewer numbers and devices
- Upload speed available at the venue
- Whether the stream will also be recorded
Reading the delay on the night
On the night, we measure the delay rather than guess it. A simple method is to display a running clock on stage, then compare it with the same clock on a test viewer watching the public stream. The difference is the real figure, and we pass it to anyone who needs to time announcements or social posts around it.
This matters for show callers as well. If a presenter says something is happening now, viewers will see it a little later, so cues to online graphics are timed against the stream rather than the room.
Knowing the number removes most of the confusion. Once everyone agrees that the stream runs, say, a few seconds behind, the delay stops being a worry and becomes one more fixed fact of the show.
Delay is built into every live stream, and the useful question is how much of it your event can sensibly carry. A longer, steadier delay suits most concerts and stage shows, while a shorter one belongs to events with a live audience taking part. If you tell us how your viewers will watch, we will set the chain accordingly and give you a measured figure before the show begins.
