Online video has been spreading like wildfire, impacting media distribution and consumption channels at unprecedented levels and providing consumers with greater control and flexibility to watch the content they want, when they want it, on the device of their choice and in high quality.
How video is delivered to these content-hungry viewers is the crux of the matter. One of the biggest challenges for online video content providers is managing the costs of content delivery networks (CDNs). Even though cost per byte has continued to decrease, the amount of bandwidth required to keep consumers happy is increasing at an even faster rate. Better networks, coverage and the growth of advanced technologies like 5G are further fueling data consumption. The significant improvements in transmission speed, quality and reliability enable close to zero latency between a streaming source and the viewer.
Yet, as hardware manufacturers create devices with bigger and better screens and users demand higher quality, bit rates just keep going up, and as the amount of data needed to deliver content increases, new approaches to video compression are an absolute necessity. This can be generated with the implementation of three methods: multi-codec streaming, per-file encoding and per-scene adaptation.
This bandwidth-saving technology enables a player to detect the specific browser — be it Chrome, Safari, Firefox or Internet Explorer — and then use the most-efficient codec. Among the options, H.264 has a big advantage in that it covers every browser on the market. H.265, also known as HEVC), delivers around 50% bit rate savings compared to H.264, and VP9 offers a very similar bit rate savings.
But there’s a catch — none of these advanced codecs can play everywhere. This is where multi-codec streaming comes in — by encoding videos into all three codecs, and configuring a player to make informed decisions about which files to stream and to whom, it allows VP9 to be delivered to more than 50% of internet users, HEVC to about 38%, while the remaining 12% falls back to H.264. Looking further ahead, multi-codec solutions will assist early movers as they adopt the new AV1 codec — which promises even larger bandwidth reductions — as soon as it is ready for a production environment.
The process of per-title encoding involves adjusting an encoding configuration to optimize specific video titles. It leverages the fact that some videos are far less complex than others, and therefore can be encoded at lower bit rates.
Cartoons are a classic example — generally, they contain scenes with large areas of solid color at a lower complexity level, which can be compressed much more efficiently than the detailed scenes in action or sci-fi movies. Pertitle encoding relies on performing a complexity analysis on each piece of content before encoding begins. That complexity score is used to adjust the bit rate ladder by sending a new encoding profile to the encoder. The result: each video in a library is encoded in a way that best suits the content, minimizing the bit rate and greatly reducing bandwidth usage.
The per-scene adaptation technique leverages the fact that the human eye is often unable to register much of the information delivered in a video stream. In most videos, there are many scenes that can be streamed at a lower bit rate without the viewer noticing.
Per-scene adaptation involves supplementing the adaptation logic that governs the player’s behavior with an additional stream of “quality metadata” that contains information about the visual complexity of a particular segment in the video. In an adaptive streaming player with a standard configuration, the player will attempt to download a video file that fits the screen where it’s playing. A player configured for per-scene adaptation can be alerted that an upcoming segment could be played at a lower bit rate without a noticeable loss of quality.
In these cases, the player adjusts itself accordingly to reduce the bandwidth consumption — in some cases by as much as 30% or more. The quality metadata required to control this process is generated by running an analysis of each video as it is encoded by using a variety of perceptual quality measurements, meaning the encoding process is optimized for the human eye. This metadata is included in the adaptive package and streams to the player in a similar manner to subtitles or closed captions.
Delivering an excellent quality of experience is a “must have” rather than a competitive advantage, but optimizing video infrastructure — including minimizing file size and bandwidth usage — is now crucial in staying ahead of the online video game.
Multi-codec streaming, per-title encoding and per-scene adaptation are all effective methods of reducing CDN costs, can be implemented into any workflow and will be able to provide a best-in-class viewing experience.
Tristan Boyd is director of marketing at Bitmovin.