What Does Low Latency Really Mean in Video Streaming?

Get more insights like this at upcoming Streaming Media East and Streaming Media West events.

Read the complete transcript of this clip:

David Hassoun: What does low latency mean to you, and what does low latency mean to me? And even then, when we actually even finally agree that we're talking about the same thing, interpretations of what low latency goals are can be completely different and very wide-spectrum. So, we're going to try to get some goals and guidelines around some of this as well.

Latency can be seen as things like time to video play, so that's more like start-up time, but that impacts also what is often viewed as live latency. So, how close am I to real time? Hand wave test. I wave my hand on the camera, and you're watching it on your device. How many seconds later do you see that hand wave? Right? They are related, because if you take longer to start, and most likely have longer latency when you're live, but that isn't completely the whole picture.

When we look at normal streaming today, which in my book is HTTP-based, so HLS, primarily. Let's say DASH is thrown in there too, because I wish DASH was even more prevalent than it is. But HLS being the primary one that we work with all the time. Normal delivering, normal latency is 30-45 seconds-plus. That's if you're doing a pretty good job. You're going to see around that type of latency.

Reduced latency is anywhere from 18 to 5 seconds, and that's where a lot of people wish they were, but aren't. Often times, low latency, really, we count that as anything less than ten seconds. Five seconds is the typical latency for HD cable. So we're talking action on the field to seeing it on your TV screen, not going through the internet; is actually about a five-second latency from lens to screen.

So anything at that or below is considered close enough to real time. If you're beating the TV, then you're doing something really right, and you're probably doing it very dangerously. Some people consider it less than ten down to 2-3 seconds, and then real time is normally considered less than one second. But there's even terms and different standards that count that differently.

Ultimately, what we're trying to say is, there's no real set information, but there's these kind of bucket guidelines. Where we're at today, and where we want to get better. How are we going to start to achieve that?