MediaKind has developed countless efficiency improvements in compression technology and entire video workflows; it’s an area we have pioneered time and time again throughout the history of our organization. Just this year, our latest achievements were recognized when we won the Technology and Engineering Emmy Award for ‘AI/Optimization for Video Compression: Improving perpetual quality metrics and their application to real-time and non-real-time video compression.’ As our CEO, Matt McConnell, remarked at the time, the Emmy award win is “the culmination of years of research and development to bring practical and innovative technologies to the market, revolutionizing the way the industry delivers television to billions of households around the world.”
In the world of video compression, everyone has one goal in mind: achieving the highest possible video quality, at the lowest possible bitrates, and all locked within a defined compute budget. Striking the perfect balance between processing power and bitrate efficiency is sound logic, of course, and is key to delivering the best possible experience to consumers across multiple devices and locations. However, there are countless permutations for how compression tools can be used by an encoder, which means the decisions made in the encoding process are a lot more complicated in practice. Traditionally, profiles have been used to offer a selection of the bitrate efficiency vs. compute tradeoff.
Video Compression: Compute performance vs. bitrate efficiency
Our dedicated in-house compression research team at MediaKind uses a strong science-based approach to drive down bitrate use, which has ensured that we remain at the apex of achieving encoder efficiency. As I like to remind our friends in the media industry, video encoders are not all equal. An enormous amount of work goes into improving the encoding process. The standards that our industry has achieved over more than 25 years – spanning MPEG-2 through to VVC – are proof that improvements and efficiency gains never stand still, with enhancements still being possible to encoders for codecs that have been around for some time.
Thanks to encoder implementation improvements, we have seen significant efficiency gains in all major encoding standards, which has meant we have continually avoided the asymptote where these codecs reach the very limit of their performance. However, the approach of traditional encoding has been one based on pre-defined profiles, with limited adaptation for how processing power is applied to the different encoding tools. Although there are myriad ways to combine different encoding methods to achieve compression, each method offers a particular benefit and cost depending on the content type. Traditional techniques have not carried the ability to leverage these various encoding methods dynamically and achieve optimal applicability.
Ultimately, video compression boils down to how we make the best use of the available resources we have. Thanks to our new AI-Based Compression Technology (ACT) development, MediaKind is once again breaking new ground in this space.
What is MediaKind’s AI-Based Compression Technology (ACT)?
As our good friend MediaKind Max explains in the video above, ACT is a new intelligent compression algorithm that enables significant operational and performance improvements. Using artificial intelligence (AI), we have developed a better way to analyze the nature of content that arrives at the encoder in real-time and drive dynamically the processing resource allocation towards different tools. AI is used to select the most effective balance of compression methods based on the nature of the content and map how the encoder uses its processing resources accordingly. This is possible for both traditional linear and ABR encoding and transcoding.
ACT has now been implemented within both our Aquila Streaming and Aquila Broadcast solutions and the latest version of our MediaKind Encoding Live product portfolio. In doing so, we can now enable encoding techniques that would have been previously too costly to implement utilizing more traditional methods. There are considerable benefits:
- Ensures optimal processing: Uses a better, dynamic choice of processing for groups of tools, producing lower bitrates for a given amount of processing resource
- Improves density: The algorithm adapts to the available resource
- Provides greater flexibility: New channels can be added without the need for new infrastructure or reconfiguring existing services
- Simplifies operations and infrastructure sizing: Removes need to select profiles ahead of time
- Enhances cloud resource management: Cloud-based deployments can set a specific, but configurable CPU limit for application deployment
In short, ACT offers an exciting and innovative way to improve encoder performance and increase compute resource efficiency, all the while reducing delivery and infrastructure costs by minimizing bandwidth. To learn more, download our new application paper.