Kudos to Bitmovin for pulling out all the stops during a fully-stocked three-week virtual extravaganza, as the encoding vendor entered its final week of NAB webinars. Bitmovin’s marathon wouldn’t be complete of course without getting AWS in for a showstopper, although it did seem strange to see the cloud behemoth so full of praise for Bitmovin over its own Elemental encoding division.
During the ‘Saving Bits with Bitmovin on AWS’ webinar, AWS was buttering up its Snow Family for data collection and migration. Snow is a great way to get content into S3 for storage, via a number of physical devices and capacity points, most with built-in computing capabilities designed to physically transport vast amounts of data into and out of AWS. Such infrastructure offload is used for scenarios such as data center shutdowns and exabyte-scale digital media migrations – which we suspect AWS is expecting a surge of once lockdown restrictions are eased, allowing physical sites to reopen and engineers to return.
“Once we get content in, we can transcode with Bitmovin cost effectively. Bitmovin can scale out transcode jobs,” said Zach Willner, Senior Parter Solutions Architect for Media & Entertainment at AWS. He noted this was a “fantastic use case” for AWS EC2 Spot Instances, which offer massive discounts for unused EC2 capacity in the AWS cloud. In this case, using Bitmovin for per-title encoding translates to a significant reduction in costs for transcode, storage and delivery, according to Willner.
“Before AWS, you would have to build huge transcoding farms to get loads of content transcoded quickly, with large upfront costs etc. AWS EC2 Spot Instances do more for less, with an average 70% off on-demand (up to 90%) price, while increasing scale and size of instances,” Willner plugged.
Bitmovin’s Principal Sales Engineer Paul MacDougall also provided his two cents’ worth, focusing primarily on per-title encoding given how this works so harmoniously with AWS EC2, while warning that companies are still being highly wasteful in terms of bitrates – translating to inefficiencies particularly in CDN and storage.
“Per-title takes the notion that no two pieces of content are the same, even though type of content varies wildly. With animation or news, for example, these are simple to compress at low bitrates – with little motion, few visual complexities and little scene by scene changes,” outlined MacDougall.
“Per-title computes an optimized ABR ladder for a given source asset to maximize QoE. We can compute a special purpose bitrate ladder based on complexity analysis, which improves quality and reduces bitrate for less complex content,” he continued.
MacDougall highlighted how a standard profile wastes a lot of bits. An example case study showed Bitmovin saving around 80% bitrate from a 1920×1080 image @ 680Kbps, compared to 3,700Kbps data to deliver the exact same end user experience.
He went on to look at the multi-codec scene. The latest data from Bitmovin shows that 100% of customers are using AVC, while 79% are now using VP9, 69% HEVC, and 22% AV1. Bitmovin has seen a significant increase in VP9 usage among its average audience, with gains in HEVC and AV1 too – advanced codecs with the potential to increase efficiencies by 40% to 50%.
There are downsides of fragmentation though, with HEVC not addressable on Android OS and Android TV devices, while VP9 is not addressable on iOS and tvOS.
“We are seeing an increase in the addressable mobile market on AV1 as more handsets roll out with hardware decoding. Previously, you needed software decoding which is very battery inefficient and effectively turns the phone into a hand warmer with the amount of heat it gives off,” joked MacDougall.
“Now with hardware acceleration for these codecs, they are unlikely to impact on processing power. Per-title encoding would not impact that, if anything this is likely to reduce CPU requirements on mobile devices as delivery is at a lower bitrate,” he added.