Discovery is using AWS as preferred cloud and ML provider to oversee its digital transformation and customise viewing experiences across its 450 TV channels and discovery+ VOD service.
Amazon Web Services (AWS) and Discovery have extended their relationship to support Discovery’s ongoing move to the cloud. Working as Discovery’s preferred cloud provider, AWS hosts most of the infrastructure the company needs to deliver the discovery+ on-demand streaming video service, its digital services and more than 450 scheduled linear TV channels to viewers around the world.
Standardising 450 Channels
Discovery’s linear channels include Animal Planet, Food Network, HGTV, OWN, TLC and Travel Channel, all together producing about 8,000 hours of original live and on demand programming per year. For program delivery, Discovery uses some of the AWS Media & Entertainment Services. AWS Elemental MediaConnect transport service for live video, AWS Elemental MediaLive and AWS Elemental MediaPackage manage Discovery’s video transport, preparation and processing.
The goal is a consistent audience experience regardless of viewing device type or connectivity. MediaLive encodes video streams for delivery to broadcast televisions and internet-connected multiscreen devices like connected TVs, tablets, smart phones and set-top boxes. MediaPackage formats video for devices and allows users to implement DVR-style viewing features like start-over, pause, rewind and so on. It can also protect content using Digital Rights Management (DRM) and scales automatically in response to load, instead of waiting for users to predict the capacity in advance.
Personalisation on discovery+
Discovery is also using AWS to increase the level of personalisation on discovery+, which features material pulled from Discovery's main channel brands’ archives, plus original series and various acquired content. It is the largest media and entertainment user of Amazon Personalize, AWS’s machine learning service for developers who want to create scalable, real-time, personalised user experiences. Amazon Personalize exposes curated content to viewers that matches their content consumption history and recommends new content based on their apparent interests, in order to customise the browsing experience.
Amazon Transcribe, automated speech-to-text conversion, is applied across both the linear broadcast channels and the streaming services to generate closed captions for hearing-impaired viewers, and Amazon Rekognition image and video analysis to automatically extract and tag video content with metadata. These tools make a searchable video library and further personalised recommendations available to viewers.
Curated content from material pulled from Discovery's main channel brands’ archives, plus original series and various acquired content.
AWS packages these services with the global infrastructure and performance Discovery requires to deliver video content, at a chosen quality and at scale, to globally distributed viewers. Avi Saxena, CTO of Direct-to-Consumer at Discovery said, “Running on AWS, we have the agility and elasticity to deliver competitive viewing experiences for peak events like the Olympics. At the individual consumer end of our business are scalable AWS machine learning capabilities that support personalisation for viewers and increase the value of their subscriptions.”
Intelligent Olympic Coverage
In mid-2021, as the Home of the Olympics in Europe, Discovery elastically scaled its use of AWS Media Services to transmit 1.3 billion minutes of the Olympic Games Tokyo 2020 content across its digital platforms, including discovery+ and Eurosport subscription services. This required delivering sports content in 19 languages across 50 markets in Europe.
Discovery used Amazon SageMaker, a machine learning service that developers and data scientists can use to build, train and deploy machine learning models for themselves, quickly in the cloud and at the edge. In this case they developed and trained models that automate the process of language monitoring. They identified the program in live broadcast video feeds and verified that the content was supplied in the correct local language for each television market.
SageMaker has tools for model training, including code-free interfaces for people who are not data scientists, and also a number of ways to make the process quicker, less costly and scalable such as automated and standardised data labelling and machine learning Ops practices, use of GPUs, and minimal latency due to inference processing.
New Fan Experiences for Discovery Sports Events
Therefore, as well as preferred cloud provider, AWS is now the Cloud, AI, ML and Deep Learning Provider for Discovery Sports Events. Under this relationship, the two companies are building an immersive fan-engagement experience for the new Union Cycliste Internationale (UCI) Track Champions League to introduce track cycling to new audiences and expand the sport’s global fan base.
Discovery Sports Events and AWS will deliver live data, analysis and statistics such as bike speed, race position and pedal cadence – that is, revolutions per minute – to fans in the velodrome where races are held, through live TV and streaming, and via a new UCI Track Champions League app. Ideally, the app will encourage fan communities to grow around the championship and help newcomers to the sport better understand the action on the track.
Discovery Sports Events will use still other AWS services to increase the immersive, real-time nature of the UCI fan experience. Amazon Kinesis, for instance, which collects, processes and analyses video and data streams in real time, will be used to ingest and process live streaming data from sensors on the track, bikes and riders, and will apply AWS analytics and machine learning to present statistics such as the riders’ biometrics, time keeping, power output in watts and so on to bring fans closer to the action.
Finally, Discovery Sports Events will again use AWS Media Services to deliver the stats to fans through the app and broadcast in a simple and compelling format. aws.amazon.com
