As you probably know, Netflix is doing pretty well when it comes to the streaming content market. It recently passed the 93m subscriber mark, it’s got upcoming movies starring Brad Pitt and Will Smith in the works, and of course there are new series of House of Cards and Stranger Things on the way.
As a result, the company is set to spend US$6bn (RM26.68bn) on content this year alone. But the company is also set to spend US$1bn on the science and technology of delivering all that content.
We met with Todd Yellin, Vice President of Product Innovation at Netflix, who explained that it’s all to do with keeping subscribers: the better the streaming experience, the more content people will watch, and the more likely they are to keep subscribing.
The first part of that initiative was to deliver Netflix’s physical servers to internet service providers themselves. This allows Netflix’s content to be as close to the end consumer as possible, meaning the content is delivered as quickly and efficiently as possible.
While Netflix holds a vast library of content, it can still hold it all on one of these server racks in ten different formats depending on the end user’s connection speed. (Netflix also asked us to put a famous movie quote on one of these server racks to give one of their engineers a chuckle when it makes it to an ISP.)
One of the challenges Netflix has is that more and more people are consuming content on their mobile.
Television viewing still accounts for two-thirds of Netflix content consumption right now, but there are three countries where more content is consumed on mobile than any other type of device: Japan, South Korea and India.
For that reason, Netflix is getting more and more serious about mobile. Currently, content is encoded at 10 different levels, and which version is served depends on the end user’s bandwidth. There’s also a setting within the Netflix app to limit the amount of bandwidth used, which is particularly useful for those people who have mobile data usage caps.
The next step, according to Yellin, is to make sure people get an even better experience when they’re viewing over a mobile network or a restricted Wi-Fi connection (such as at a hotel). The answer is more advanced encoding techniques, and adaptive encoding that can utilise less data for less detailed visual scenes.
The first part of the puzzle is just getting more efficient at encoding, meaning you can do more with the limited bandwidth that can often apply to users on mobile networks. As anyone who’s watched the HBO series Silicon Valley will know, compression algorithms are kind of the holy grail of multimedia delivery companies: the more you can compress, the less data you use. And as data costs money, both from the end user’s perspective and the internet service provider’s, everyone’s kept happy.
Yellin showed us a demo of Stranger Things on two identical mobile devices, and asked me to pick which one I thought was better. The one on the left was much worse - all blocky and full of artefacts, while the one on the right was a much cleaner picture. Yellin revealed that they were both being encoded at 100kbps (a very low bitrate), but that the one on the right was using a new and improved codec that will be released by Netflix in the coming months.
It’s difficult to tell from these images, but the version on the right is much cleaner. It’s based on Google’s open-source VP9 codec, which allows for a more detailed picture at the same bitrates used in previous encoding techniques.
However, VP9 also allows Netflix to be as efficient at lower bitrates as it currently is at higher ones. For example, we were shown two trailers for the upcoming Marvel’s Iron Fist, and were asked to spot the difference. This time it was much more difficult to tell - they both looked free of artefacts and we’d be happy watching them on-the-go.
The reveal was that one clip was encoded at 555kbps in the current codec, while the other was encoded at 277kbps in the new VP9 codec. The result is that Netflix is able to produce the same picture quality at half the bitrate.