WATCH THE FULL VERSION OF THIS MASTERCLASS NOW
sides. They’re much more agile partnerships and are doing more with less. You often have a smaller audience for more channels. So the economics are different than they used to be. We are trying to give our broadcast operations customers tools to visualise that entire workflow – all the way from the encoder, through to distribution, as well as being able to manage things like failover in the cloud. We use AI and ML in our toolset, so broadcast operations engineers can see the patterns that are repeatable, that we observe across customers. We can say: ‘there’s a pattern here that indicates something’s not quite right’, and give them a heads up on failures.
We have models that are trained with something like three billion data points every day. So we’re able to see across the industry what the sources of instabilities are in these complex ad hoc workflows, and where problems are coming up. We hear customers talk about spending hundreds of thousands of dollars a year, per channel, just to figure out what happened on these failures. BILL ADMANS: For us, the supply chain that’s being used by large studios, networks and content owners, often involves petabytes of content and thousands, if not millions, of files. That’s humanly impossible to manage on your own. So, we use ML and AI to manage
their content and orchestrate how it is being used through the workflow. It also helps our customers manage the cost of that media in the cloud. One of the areas where AI really pays off is in the deduplication workflow, where you might have many versions of titles building up over time. A typical Netflix show could have 26 different versions of a single episode, because of the different languages being ordered. Yet, in reality, 90% of the frames are identical. When you add up the size of a video frame being stored over and over again – and we all know how massive video files are – it gets very expensive, very quickly. Then, multiply that over thousands, or tens of thousands, of shows in a library.
@feedzinesocial
Powered by FlippingBook