In a move that highlights the value of media optimization, online streaming giant Netflix announced that it is experimenting with ways to reduce bandwidth, while providing the same (or better) viewing experience.  Netflix has it right, there are simply too many wasted bits being transmitted by video services today.  As pioneers in the field of media optimization, Beamr is energized to see the world’s top OTT content distributor acknowledge the tremendous need to optimize their video, and we welcome the increased attention Netflix will bring to this important issue and its impact on the video distribution ecosystem.  But the question is who will follow their lead?

The fact that Netflix cares about reducing bit-rate is a great thing for the Internet and all consumers, whether subscribers of the video service or not.  Sandvine, the global bandwidth measurement firm, reported in their most recent Internet Phenomena report that during peak hours Netflix traffic occupied a whopping 37.1% of all global traffic.  Which to put into perspective, means Netflix represents more than double the traffic from Google’s YouTube, which during the same period generates 17.8% of all downstream traffic.  This means combined Netflix and YouTube comprise greater than 50% of all the bits flowing over the Internet during peak hours.

But how Netflix will achieve this reduction in bandwidth without negatively affecting video quality–this is the $64,000 question–let’s take a look.

Savings + quality improvements = a win for the Internet and consumers

Contrary to popular thinking, current encoders have not yet reached the state of the art when it comes to matching optimum bit-rate to content.  Which explains why Netflix felt the need to innovate.  The secret to their solution is based on not using a fixed set of resolutions and bit-rates across the library (as is standard practice today), but rather pre-analyzing every title, setting the encode bitrates and resolutions based on the need of the video content, and taking into consideration factors such as temporal motion and spatial texture.  With this novel approach, Netflix is able to reduce their average bit-rates and at the same time improve perceived quality across their library.

Why are we so bullish that this is the right approach?  Because it turns out to be similar to what we’ve been using in our own commercial solution for more than two years now.  

As Netflix states in the blog post that describes their solution, most adaptive bit-rate (ABR) encoding recipes are fixed, and historically ABR settings have been prescribed by Apple as “best practices” for HTTP Live Streaming (HLS) in Technical Note TN2224.  This is the reference many video engineers quote when quizzed on how they arrived at the bit-rates, resolutions and parameters selected for their recipes.  However, these recommendations were made more than ten years ago when the state of both video encoders and network bandwidth was at a very different place than it is today.  Put simply, a video engineer could follow Apple’s “recommendations” precisely, only to discover that the resulting video quality is inconsistent or subpar across the library.  Netflix seems to have arrived at this same conclusion with their move away from fixed recipes to content adaptive ones.

Breaking out of legacy constraints

In the early days of video–when the spec for how a stream should be built was not much more than a loose set of guidelines–in order to reliably deliver a quality entertainment experience, many variants of a file had to be built using different codecs, bit-rates, encoder profiles, etc.  This means many services encoding recipes have evolved as a result of legacy constraints that over time forced decisions resulting in video quality or bit-rate compromises.  One area where this situation can be observed is with the high number of services still using constant bit-rate (CBR) even though variable bit-rate (VBR) encoding is able to yield better quality per bit-rate ratios than CBR.

Even still, today, most encoding recipes are fixed, which means the same set of resolutions and bit-rates are used across all titles, even though the quality of video at the same bit-rate will change from title to title based on the specific (actual) needs of the video content.  There are many resources to help explain this further, but a great place to start is the Netflix technical blog.

How can the average OTT streaming provider achieve equal or better results without an army of engineers and imaging experts?  

Beamr offers a commercial product that performs content-adaptive media optimization starting from any fixed bit-rate encoding recipe.  Available in two versions: on-premises and as a fully managed cloud service, Beamr Video selects the optimum bit-rate needed per resolution to achieve the best streaming quality for each movie regardless of content type—flat animation, 3D animation, action, drama, etc.  A unique feature of Beamr’s content-adaptive optimization is that the product modifies bit-rate for every scene at the frame level using a closed loop process controlled by a patented perceptual quality measure.  This method allows unparalleled aggregate savings while fully preserving the original encode quality.  For more details on closed-loop perceptual quality optimization, see our blog post on this topic.

How Netflix and Beamr approaches converge

Consider a movie with both high-motion action scenes, and low-motion talking head scenes or scenery.  If the encoder allocates a fixed bit-rate based on the “average” characteristics of the video, the talking head scene will get more bits than needed, wasting precious bandwidth–while the action scenes may not get enough bits allocated–resulting in visible artifacts and low quality video.  Only by adapting the optimum bit-rate based on content type, can a video service achieve consistent quality across the entire library without throwing away bits.  Though Netflix’s approach is to identify the needs of the video per title, while Beamr takes the process down to the frame level and per scene, both companies’ approaches are centered around a quality measure driven analysis process, which can move the current state of the art forward for video encoding when examined thru the lens of quality vs. bit-rate.

Thank you for joining us Netflix in rethinking how efficient encoding can be achieved while contributing to the “de-cluttering” of the Internet.

If you’re intrigued, please visit www.beamr.com to see video samples and learn more about Beamr’s technology, solutions and customers.


Share