Bitrate, Keyframes and Latency: Three Streaming Terms Explained Plainly
A streaming engineer explains bitrate, keyframe interval and latency in plain words, and what each one means for the audience watching your live event on a phone or laptop.
Why these three words come up on every call
When we plan a live stream with an organiser, the same three words appear in almost every conversation: bitrate, keyframe interval and latency. They sound technical, yet each one answers a question an organiser already has. How good will the picture look? How quickly will viewers see what happens on stage? How likely is it that the stream stumbles halfway through the headline act?
I look after the streaming chain at entertoly, from the encoder in the gallery to the platform that carries the event to viewers. Over the years I have found that people make better decisions once the vocabulary is clear, because the trade-offs stop being mysterious. Nothing here needs an engineering background.
So this note takes the three terms one at a time, says what they mean in everyday language, and shows where they pull against each other. It is about the words themselves, not a pitch for any particular set-up.
- Bitrate: how much data per second you send
- Keyframe interval: how often a full picture is sent
- Latency: the delay between stage and screen
Bitrate: the amount of picture you send each second
Bitrate is the quantity of data the encoder sends out every second, usually counted in megabits. Think of it as the width of a pipe. A wider pipe lets more picture detail through, so fast movement, stage lighting and crowd shots hold together better. A narrow pipe forces the encoder to throw detail away, and you see it as blockiness or smearing, especially when the lights flash.
A higher bitrate is not automatically better. It has to fit through your upload connection with room to spare, and it has to suit the screens your audience is using. Someone watching on a phone over mobile data gains very little from a bitrate chosen for a large television.
We normally set the bitrate well below the measured upload speed of the venue, leaving a clear margin. A stream that runs at three quarters of the available capacity is far more forgiving than one squeezed right up to the limit.
Keyframes: the full pictures between the savings
Video is not sent as a stack of complete photographs. To save data, the encoder sends a full picture now and then, called a keyframe, and between them sends only what has changed. The keyframe interval is the gap between those full pictures, normally given in seconds.
Why does it matter to you? Viewers who join late, or whose connection hiccups, can only recover cleanly at the next keyframe. A short interval means a quick recovery and a quicker start when someone opens the stream. A long interval saves some data but leaves people looking at a frozen or broken picture for longer.
Most streaming platforms ask for a particular interval, and a mismatch can cause odd stuttering or a rejected feed. Checking this setting against the platform's published requirements is one of the first jobs when we build the chain, long before the cameras are powered up.
- Short interval: faster recovery, slightly more data
- Long interval: less data, slower recovery for late joiners
Latency: how far behind the stage the audience is
Latency is the delay between something happening in the room and a viewer seeing it. Every step adds a little: the encoder, the journey to the platform, the platform's own processing, and the buffer on the viewer's device. A stream that is twenty or thirty seconds behind is perfectly normal, and for many concerts it causes no trouble at all.
It becomes a problem when the audience is meant to take part. If viewers are voting, asking questions or following a countdown, a long delay makes the interaction feel out of step. Lower latency settings exist, but they reduce the cushion the stream relies on to survive a wobbly connection.
This is the clearest example of a trade-off. Faster delivery and a safety margin pull in opposite directions, so the right setting depends on what your event needs the audience to do, not on a figure that suits every show.
How the three terms pull against each other
Raise the bitrate and the picture improves, but the stream needs more capacity and is less tolerant of a poor connection. Shorten the keyframe interval and recovery speeds up, yet each second costs a little more data. Cut latency and the audience feels closer to the stage, but you lose some of the buffering that hides small problems.
None of these settings can be set in isolation. A change to one usually asks for a second look at the others, which is why we settle them together during planning rather than adjusting them on the night. A late change to a single value, made under pressure, tends to cause the very trouble it was meant to cure.
What I aim for is a chain that is unremarkable. Viewers should not notice the encoder at all, and the organiser should hear nothing from the audience except that the show looked good.
- Agree the audience's devices and connections first
- Set bitrate with a margin below the venue's upload speed
- Match keyframe interval to the platform's requirements
- Decide how much delay the event can live with
Questions worth asking your streaming team
You do not need to memorise any of this, but a few plain questions will tell you whether your streaming team has thought it through. Ask what bitrate they plan to use and how it compares with the venue's tested upload speed. Ask what happens if the main connection drops, and whether a second route is available.
Ask how they will know the stream is healthy while the show is running. A dedicated person watching the outgoing feed, rather than someone glancing at it between other jobs, is what lets a problem be caught within seconds instead of after the audience has complained.
Finally, ask for the platform settings in writing before the day. If the numbers are agreed in advance, nobody has to guess in the middle of a soundcheck, and the engineer can spend the evening watching the stream rather than rebuilding it.
Bitrate, keyframe interval and latency are three dials on the same panel, and each one trades something for something else. Knowing what they mean lets you state what your event needs, whether that is the cleanest picture, the quickest delivery or the safest margin. Bring those priorities to the planning conversation and we can set the streaming chain to match them, then watch it closely for the whole of the show.
Daniel Foster
Streaming EngineerDaniel builds and monitors the streaming chain so events reach audiences without interruption.