The State of CMAF: The Holy Grail or Just Another Format?
First, understand that CMAF requires separate audio and video tracks. If you’ve stored all of your content in muxed file formats, you’ll have to reprocess that content to work with CMAF. In his NAB presentation, NBC’s McLary commented on this issue: “Most HLS that you work with has the audio muxed in. Not 100%, but practically that’s what’s out there. So as you look for ways to start to mix your HLS and your CMAF workflows together, audio becomes a big deal, especially when you deal with things like server side advertising insertion [SSAI], which becomes really complicated really quickly when you have to deal with demuxed audio.”
When it comes to advertising insertion, many AVOD services don’t control their own destiny. Discussing the conversion to CMAF, one large premium content publisher, who asked not to be named, said, “We’re ad-supported, so we’ve been waiting on a few things to fall into place before we can pull the trigger. First is Google Ad Manager’s support of CMAF, which we understand to be coming by [the end of the year]. They are our main decisioner and SSAI stitcher, so support here is crucial. Particularly on the stitching side where I believe the demuxed audio requirement is the most challenging for them at the moment. They’ll need to ensure all ad creatives meet CMAF specs.
“We are also working with our packaging vendor to be prepared for the specs, which Google will require to even start packaging CMAF. Which we currently don’t support. So ideally, we can get our ducks in a row to do CMAF when Google gets their ducks in a row. Ad insertion has always been the tech which slows us down. Always.”
It’s not just advertising insertion. As THEO Technology’s Speelmans says, “Depending on the metrics you have, you could have a need to update your analytics and monitoring pipeline.” In this regard, I asked Matthew Driscoll, product management director for Telestream iQ, about CMAF support, and he responded, “IQ’s monitoring probes support ISOBMFF for both HLS and DASH streaming protocols. Once the CMAF media is referenced in either the HLS playlist or DASH manifest, it can be actively monitored either post origin or post cache.”
In terms of how long it should take to convert over to CMAF, Speelmans says, “In my experience, most packagers already support it today, so it’s only a matter of reconfiguring. Players are usually able to support it transparently as well. It will depend on the setup of your pipeline to get this done, but I would assume no more than two weeks of work. Keep in mind you would want to do a full test afterward, meaning you would need some additional time for that.”
Of course, as Eyevinn’s Svensson points out, “The more functionality you add, the more complex it becomes. DRM and ad insertion is complex, even without CMAF. I’m not sure if CMAF itself adds more complexity, it’s more the combination of formats, device support, and features. If you want to reach all kinds of devices with DRM-protected content and at the same time do dynamic advertising, it becomes complex.”
It’s More About Simplicity Than Low Latency
Low-latency CMAF has emerged as a viable technology to decrease latency down to 1–3 seconds, depending on whom you ask. Still, several of the OTT vendors currently implementing CMAF warn not to equate CMAF with low latency. One, who wished not to be named, says, “It’s not that CMAF equals low latency. It’s just that everyone is focusing on both so they conflate.” Another, who also wished not to be named, says, “Start switching to CMAF now; between the Apple Low Latency HLS spec and other CMAF-based approaches, it’s going to take a while for the low-latency side to sort out. Don’t delay your CMAF implementation for that.”
What emerged as the most pressing motivation to adopt CMAF for most OTT shops was simplicity, not low latency. At NAB, Cooper Pope, director of multiplatform video solutions at WarnerMedia, showed Figure 4 and commented, “I can think of six different ways that we’ve implemented closed captions. I can think of four different thumbnail scrub previews. I can think of a half dozen or more methods for ad insertion. And whenever you add a new device, it just adds to the list of busywork you have to do to maintain feature parity. At this point, you aren’t innovating, you’re just replicating what you’ve already done, and so, you’re looking for a better way to do things right.”
WarnerMedia hopes to simplify content delivery with CMAF.
Another OTT vendor says, “The other obviously huge driver for us is unburdening ourselves from fragmentation. We generate four different packages for DRM support across platforms. (I’m sure other publishers are customizing packages for mobile/CTV/web as well). The idea of moving to CMAF/CENC is quite appealing to us. The benefits of fewer packages on encode/packaging/storage costs are also not lost.”
Anevia’s Blanc effectively summarizes the concept: “The key potential benefit of CMAF that few people talk about is simplicity. Some customers are currently delivering six different combinations of ABR format and DRM, some even more, which is incredibly complex in terms of testing and [quality control]. If they could send one format to all devices, CMAF would deliver huge cost savings in terms of reduced complexity and reduction in test requirements.”
CMAF Is the Next Big Thing
Imagine how television would have evolved if every TV manufacturer had to test its new sets with every channel on the planet, and every channel had to test with every new television set from all manufacturers. Incompatibilities would have run rampant, and market growth would have been stultified. In essence, that’s what’s happened in the streaming space, which pushed a significant compatibility burden on OTT publishers. Given this dynamic, it’s amazing the OTT industry has been as successful as it has been.
In essence, this compatibility issue is at the heart of what WAVE is designed to resolve. Technically, WAVE is a project hosted by the Consumer Technology Association (CTA). CTA’s website says that the project’s goal is “to improve how internet-delivered commercial video is handled on consumer electronics devices and to make it easier for content creators to distribute video to those devices.”
I spoke with Will Law, chairman of the technical working group, and Microsoft’s John Simmons, a working group member who helped design web standards like the Media Source Extensions (MSE) and Encrypted Media Extensions (EME) and worked with Apple to develop CMAF. The pair pointed out that the WAVE project was founded right around the time CMAF was standardized and now has more than 60 members throughout the streaming ecosystem (see a complete list of project participants).
WAVE plans to promote interoperability by creating specs and test suites for content, devices, and APIs, which previously haven’t existed. Not surprisingly, CMAF is at the heart of all the specifications. At IBC 2019, WAVE launched the CMAF Industry Forum, chaired by Apple’s Krasimir Kolarov. The forum is an offshoot of WAVE designed to highlight CMAF’s role in the WAVE specs and test suites and to encourage adoption and usage (Figure 5).
The net-net is this: CMAF was developed by Apple and Microsoft as a unifying container for multiple ABR formats, including HLS, DASH, HLS, and HDS. The WAVE project is focused on using CMAF to create specs and test suites to ensure content/device interoperability. In 2 years, content publishers won’t choose an encoder/packager that hasn’t passed the appropriate test suite, and no player, hardware, or software will ship without similar testing. New features, APIs, and codecs will be added in a standardized way, allowing true innovation rather than just the busywork of making the content play on the target device.
By buying into CMAF, you’re not just buying into a new container format—you’re buying into an industry organization with the vision and clout to turn a simple spec into interoperability. It won’t happen in the short term, but as one publisher says, “Give it a few more years for AV1, and we can maybe put the pen down on this thing.” There’s no other standard or specification that comes close to delivering that kind of vision.
CMAF Isn’t for Everyone—Yet
All that said, CMAF is not for everyone, at least not yet. When I asked Steve Heffernan, cofounder and head of product at encoder vendor/ OVP Mux, about CMAF, he said, “Today we don’t use CMAF, only HLS+TS. Our customers rely on us to make format decisions for them based on the features they need, and we chose to start with HLS+TS because it has had the broadest single-format reach, including older iOS devices. We’ll likely switch over to CMAF delivery in the near future, thanks to iOS CMAF penetration, and building in DRM support.”
Marcus Johansson, a streaming engineer at OSK Berlin, says, “We haven’t really implemented CMAF in any projects because our current platform already works with HLS on all the devices and browsers that we need to support. Ultra-low latency live streaming has also not been a requirement this far. So we had no need for sharing distributed resources over DASH/HLS or using chunked streams, though we’re starting to get some inquiries and have low-latency prototypes running in the lab.”
Greg Ellis, COO and VP of business development and sales at DaCast, says, “Everyone wants the lower latency, and CMAF looks like the best option with real scalability. That said, demand from our bigger and fastest-growing customers for other enhancements with higher priority is causing us to push CMAF off almost every quarter this year.”
So we’re not hearing a lot of “no,” just a lot of “not yet.”
Most of the Industry Is Gearing Up
Otherwise, most technology companies I spoke with have either implemented CMAF or are moving full speed ahead. They include encoding vendors like Bitmovin, Brightcove, Capella Systems, Encoding.com, Hybrik, Media Excel, MediaKind, Mux, Telestream, and Videon Central. Encoding.com is even adding a CMAF compliance check to its quality-control process to ensure compliance with the specification.
CMAF is fully supported by most player vendors, including Bitmovin, JW Player, and THEO Technologies. Akamai has supported CMAF for more than a year and a half, while Limelight Networks has proofs of concept working and plans to launch in 2020. Cloud transcoding vendors Wowza and Softvelum support CMAF now.
All the consultants I spoke with have at least one CMAF project under their belts. In addition to those mentioned above, RealEyes CEO David Hassoun reports that he helped one unnamed OTT vendor “migrate from HLS transport Streams to CMAF. The primary goal was to unify DRM across the board especially for the Web in part to replace Flash (Flash still used for DRM only).”
Consultant Mark Kogan reports helping one large Israeli telecom start up a CMAF-based 4K HEVC HDR service to stream the World Cup to Apple TV clients. The service is now being expanded to encode DASH targets like LG, Samsung, and other connected TVs.
As shown in Figure 6, 25% of participants in Bitmovin’s “Video Developer Report 2019” are planning to implement CMAF in 2020. Given that 41% identified “playback on all devices” as their biggest challenge (with latency first, at 54%), this is hardly surprising.
25% of participants in Bitmovin’s “Video Developer Report 2019” are planning to implement CMAF.
Basically, everywhere I looked, CMAF was either in use, under development, or strongly under consideration. It’s not too late to start working with CMAF, but it’s certainly not too early.
[This article appears in the October 2019 issue of Streaming Media Magazine as "The State of CMAF."]
CMAF has reached maturity. Here's one way to go about using it, with AWS Elemental MediaConvert
Here are two strong reasons to attend Streaming Media West's pre-conference training sessions: a morning seminar on deploying CMAF and an afternoon session on video machine learning.
Microsoft's David Sayed, Imagine Communications' David Heppe, and Akamai's Will Law discuss CMAF and the future of packaging formats in this clip from their panel at Streaming Media East 2019.
RealEyes Media Development Manager John Gainfort discusses CDN scaling, chunked encoding, and their future impact on latency in this clip from his Video Engineering Summit presentation at Streaming Media East 2019.
Reducing latency for HTTP Adaptive Streaming video to 3 seconds or less is possible, but it requires a complex workflow.
Companies and Suppliers Mentioned