Contributory BlogsOpen Source

Open Source Could Give Game of Thrones Multiple Endings

0

Intel’s open-source technologies make it possible for creators to be more innovative and create shows like Black Mirror: Bandersnatch

The final season of Game of Thrones has begun and we are slowly inching towards its ending, one week at a time. “I think not everyone’s going to be happy, you know, and you can’t please everyone,” said, Kit Harrington. GoT showrunners can please everyone if they do a Bandersnatch.

One ending could be with Denny as Queen and another one with Jon Snow as King. What if Cersei has a change of heart or she brokers a deal with the Night King…. You could make ever viewer happy if you let them choose their own ending.

‘Choose your own adventure’ is not a new concept, but Bandersnatch brought interactive entertainment to adults. The show gave us a peek into the future of entertainment.

As challenging as it is to develop shows like these, it’s even more challenging to bring them to the audience. One of the biggest cost centers for streaming services like Netflix is bandwidth. Shows like Bandersnatch push that challenge even further as they change the traffic dynamics. The interactive nature of the show creates back and forth traffic as users make different choices and Netflix delivers results based on those choices.

That takes a great deal of bandwidth. Typically, content is stored at locations near users using content delivery networks (CDN) to speed up the delivery of content. But legacy CDNs are optimized for one-way traffic, not for interactive shows like Bandersnatch.

Bandersnatch is just the beginning of the content that’s going to put more load on bandwidth. Companies like Google and Microsoft are building cloud-based game streaming services. Microsoft has announced cloud-based virtual Windows 10 instances. All of it requires two-way traffic and new efficiencies.

New problems need new solutions. The first and foremost problem to solve is compressing ultra high definition content without compromising on quality so more data can go back and forth via existing pipes. Compression is important however; given the amount of processing required and the numerous different device types that need to be supported transcoding (encode/decode) becomes critically important. This is where efficient codecs become important.

Intel has been working with companies like Netflix to build next-generation codec, called SVT-AV1 (Scalable Video Technology for AV1), which is optimized for modern workloads.

50% more efficient than existing solutions

AV1 aka AOMedia Video 1, is a video coding format designed for royalty free video transmissions over the Internet. SVT-AV1, an AV1 compliant encoder, is much more bandwidth efficient (up to 50%) as compared to the most popular CODEC today, AVC (which is used by 60% of the industry). More than 50% efficiency is a game changer for CDNs, video content developers and streaming services.

Lynn A Comp, Vice President, Data Center Group, General Manager of the Visual Cloud Division at Intel Corporation told TFIR that with SVT-AV1, Intel is accelerating commercial viability of AV1 which will help lower the cost of delivering content. This also means companies can be more innovative without content, without having to worry about bandwidth.

Software-driven scalable solution

As a software-defined codec library, SVT-AV1 is highly scalable and offers a perfect balance between performance, latency and visual quality when working with visual cloud workloads.

SVT-AV1 codec allows encoders to scale their performance based on workloads. If Netflix benefits when delivering interactive shows, platforms like Twitch benefit when blasting screen capture scaling from one user to 50,000 users at the same time.

In addition to SVT-AV1, Intel has also announced the Open Visual Cloud, an open-source project that includes a set of use case-optimized reference pipelines for visual workloads.

These developer-ready pipelines are based on open-source media, artificial intelligence (AI) and graphics software ingredients. They support the most popular open-source frameworks that developers are familiar with.

SVT, the OpenVINO Toolkit and the Intel Rendering Framework are all part of the Open Visual Cloud, bringing highly optimized open source encode, decode, inference and graphics together as interoperable reference pipelines for services innovation around targeted visual use cases. The first two pipelines enable services for 1) content delivery network (CDN) transcode VOD streaming and 2) intelligent ad insertion.

Intel already has all the building blocks for encoder-decoder inferencing and a rendering framework. But the company never put them together as a single solution. ISVs and other players have been telling Intel that putting all these open source pieces together and optimizing them would help the industry.

Open Visual Cloud put all these open source components together and along with GStreamer -and FFmpeg plugins for optimized CPU and HW acceleration, Intel has created developer-ready reference pipelines.

Intel has open sourced SVT encoder core libraries (HEVC, VP9, and AV1) as there was duplication and wasted efforts where companies were trying to optimize the same core bits over and over without any collaboration. The industry was not innovating on top of existing solutions, it was reinventing the wheel. By bringing an optimized core, Intel not only saved countless hours of such companies it also gave them something that they can build on top of to differentiate and start moving forward more quickly.

A new era of creativity and innovation

Open Visual Cloud also solves talent and logistic problems for studios. Studios have to either move their whole team at one place when working on a project or ship expensive workstations to people. It would be a dream come true if they can leverage the visual cloud so that anyone sitting in Thailand and Africa could do animation rendering. It also means they can tap into the talent pool that they currently can’t. As a result, it will unleash a new era of creativity.