Leaping Into 2016: What Still Needs to Be Solved in Streaming
Three items for the online video industry's to-do list: Challenges for optimization, latency, and encoding that can be fixed in the short term.
Learn more about the companies mentioned in this article in the Sourcebook:
When I started in the industry back in 1998, having moved from traditional cinematography through nonlinear editing and on to videoconferencing, I wrote up a list of major issues the streaming industry needed to solve before streaming could replace television, or at least become the consumption platform of choice.
The list was fairly simple, with no more than 20 primary issues to address, but the tasks themselves were quite daunting. One, for instance, was to scale beyond television-sized audiences. As we now know, that has inspired several different ways to scale, and the latest approach— using HTTP-based dynamic streaming, in various flavors such as HLS or DASH—still leaves lots to be desired.
One area that no one seemed to touch, though, was the fundamental way that encoding quality could be improved.
Yes, we’ve had one or two iterations, but nothing revolutionary. If you think about it, H.264 was part of the videoconferencing world long before it was part of streaming. I wrote about the history of H.264 in my “Decade of Streaming Media” article in 2008 (which is not online, unfortunately), and it will probably still be central to the “Two Decades of Streaming Media” article in 2018.
H.265 is mostly a variation on a theme, as is VP9, which means that we need another way to solve the quality issue without sacrificing performance. And that will require fundamental research.
Ideally, a new set of algorithms would need to be created and fed into the encoder—at the time of encoding, not in a post-encoding process—to fundamentally alter the output of the encoder and achieve better quality at the same bitrate. Conversely, if the algorithms were good enough, the same “status quo” quality could be maintained at lower bitrates, or so the theory goes.
Which brings us to 2016, and the fact that I’m now employed by a company, EuclidIQ, that is working to solve that “better compression” issue while still maintaining standards-compliant playback in industry-standard H.264 players. (The “employed” bit represents a conscious decision to leap into the fray, moving from consultant to employee.)
So let me take this first column of 2016 (a leap year) to issue a challenge the industry to leap into the year with the intention of solving some additional issues from my list.
There are three more issues from that list that could be addressed in the near term.
First, there’s a need for codec-agnostic quality optimization. That means we’re not stuck with a single codec, such as H.264 or H.265 or even VP9, but have the flexibility to apply these learnings across a wide variety of codecs.
Second, and this is a challenge to the industry, as we all need to get back to basics and solve fundamental issues: We need to figure out a way to lower our broadcast latencies, from the delay of 70 seconds or more that we currently see for live streaming, to a number much closer to a traditional broadcast delay of 3 seconds or less.
Finally, there are several other minor issues to be addressed: object-based or model-based encoding reuse—great for indexing, search, and retrieval technological needs—as well as the need to revolutionize audio codecs so that we can get AAC+ equivalent surround sound down around 16Kbps.
Doing these things won’t be easy, but we’ve got the right circumstances coupled with the right minds to be able to solve these problems in the upcoming year. So leap in and let’s fix these things!
This article appears in the January/February 2016 issue of Streaming Media magazine as “Leaping Into 2016.”
The industry is in the early days of streaming entertainment, and there are still plenty of obstacles that need to be removed.
As 2016 begins, StreamingMedia.com looks at the content, monetization, and workplace challenges that face the industry. The more things change, the more they stay the same.