8K 60P VVC Latency Less than 1 Second
A recent post by Spin Digital on LinkedIn set out the low latency that it has achieved for encoding 8K 60fps 10 bit video to VVC and playing to an 8K TV. The company sent the encoded signal to its own Spin Player for viewing and the overall latency was just 600ms. This got us thinking about the different requirements for video latency. They vary a lot according to the application.
Traditionally, over-the-air (OTA) broadcasters would often deliberately introduce what they called a ‘profanity delay’. This could be of a number of seconds (it is also sometimes known as a ‘seven-second delay’) and was used to allow either a blanking of the video or audio or ‘bleeping’ to cover profane language or other ‘undesirable events’ – think of the famous Janet Jackson ‘wardrobe malfunction’ at the Super Bowl in 2004. Where broadcasting is legally regulated, broadcasters can find themselves in trouble with the authorities over this kind of issue.
Encoding is Processor Intensive
However, the delay can also be a result of the amount of processing and computation that is needed to produce and distribute video signals. Codecs are typically designed to be asymmetric – that is to say, they tend to need a lot of processor power at the point of encoding in order to minimise the amount of processing needed in the decode. After all, a single video stream may only need to be encoded once, but might be decoded in millions of homes or on millions of devices. (This asymmetry has an impact on many areas of codec adoption and development, but that is a topic for another day).
Early on in the development of a new codec, compression is usually done by CPUs and GPUs – and Spin Digital said that the example we quoted was based on CPU compression. However, as the codec becomes stable and better understood, special processors are usually developed to optimize the compression process. However, the market scale for encoders is much smaller than for decoders which can impact the speed of development.
Different Delays
Now, when there were only a limited number of distribution channels for video – I.e. terrestrial, then cable and DTH satellite, there was not so much difference in the latency of the video signals arriving in the home. Moving to the use of the open internet for distribution of video content, on the other hand, can show big variances in the arrival time of video. There is a good explanation of the kind of delays introduced by ‘multi-hop’ delivery provided by CDN company, Cloudflare here. CDNs are Content Delivery Networks – organizations that help to minimize the latency of distributed content by caching content close to the consumer.
Another reason that latency may be deliberately introduced is to allow a buffer to be created at the client viewing station to minimize interruptions to the video display if there are delays in the transmission process. Delays of even beyond 30 seconds have not been uncommon. The chart below is from the Bitmovin 2024 Developer Report in 2024 and it shows that half the developers at that time still had a latency of 10 seconds or more.
Different Applications
However, different video content makes different demands on the latency. While it may not matter a lot whether a feature film starts at a particular time, some content, and especially live sport can need a much lower latency. Viewers may have apps on their devices or TVs as well as the video and if the apps are reporting significant events ahead of the video, it can have a negative impact on the viewing experience. Equally, some broadcasters are trying to develop more social ways of enjoying content and that is potentially a challenge if different viewers are receiving images at different times. In this case, video providers can deliberately set a longer latency so that all the streams are displayed at the same time.
Then again, there are many applications where latency is even more critical. For example, online sports betting crucially depends on low latency and consistent performance. In this kind of case, there may well be a trade-off between the complexity of the codec and visual quality and the latency. This application has a lot of value and growth and so there are specialist operators aiming to help service providers optimize their performance.
If you are interested in a deeper dive into this topic, you might enjoy this recent podcast with codec specialist Jan Ozer and Peter Jan Spielmans of Theo Technologies. In it, Spielmans highlights that it’s not so hard to get video streaming latencies very low in small projects, but in big events and with global reach, a latency of six or eight seconds is more realistic. The podcast also covers quite a lot of information on codec adoption and usage (from around 33 minutes in.
