MPEG DASH's Future Is Fractured, Says VideoRX
MPEG DASH offers a promise of simplified endoding and delivery for adapative streaming, but that promise is still years away. Speaking in a red carpet interview at the recent Streaming Media East conference, Robert Reinhardt, inventor for VideoRX, says the problem is getting the major players to agree.
"Let's just say Apple continues with HLS and that's all they do for adaptive streaming and they are like "Eh, MPEG DASH, we don't really need it, we've got our own solution already." But then we still have this environment where we have two streaming standards," noted Reinhardt. "Two is better than what we currently have, which is Adobe, Microsoft, and Apple all sort of pushing their own adaptive streaming manifest "
The future of MPEG DASH was a hot topic at the conference, and Reinhardt likes what a fellow speaker had to say on it:
"I love the quote from Will Law at Akamai who's on the MPEG DASH panel here, he said that all the adaptive streaming specifications are 80% similar and 100% incompatible with each other," said Reinhardt. "I love that because they all pretty much do the same thing right and they do it almost the same way with the video. They chunk up the video, most of them have rallied around H.264 and it's just a different specification for how those things are wrapped up, right?, but essentially the structure is the same."
To view the full inteview, scoll down.
Troy Dreier: Hi everyone, this is Troy Dreier, senior associate editor for StreamingMedia.com coming to you from the red carpet at Streaming Media East in New York City. I am joined right now with Robert Reinhardt, inventor for encoding and hosting company, VideoRX. Since you're one of our HTML5 specialists, let's start out with a little easy question, what does HTML5 mean for video content owners?
Robert Reinhardt: I guess I could start with a broad generic case for HTML5 video and what it would mean to the vast majority of my clients is to center around H.264 and have that as a chosen acquisition, production output codec and use playback technologies that support that. HTML5, as most people know, doesn't support H.264 consistently, but where it isn't supported, we still have fallbacks to things like Flash player that can play H.264.
Troy Dreier: Now, I was going to ask you about how those who want to take advantage of HTML5 can prepare their content for it, I guess you've answered that a little bit. Are there other things people should be doing to prepare their content for HTML5?
Robert Reinhardt: If you are someone who only has a few pieces of video on your site, even if you update it routinely, like let's say it's a marketing video and that's essentially all you do with video, I am not saying that's not a lot, but if you only have a few pieces of video that you have to maintain regularly for any kind of forward-facing website, it makes all the sense in the world to encode into all the formats that HTML5 would support, right? Everything from H.264 Baseline, Main, WebM, and even Theora if you want to go crazy because you don't have a lot of content to manage, right?, and why not have it in all these formats that is going to enable potentially an outlier, right? Because when we talk about audiences, we are mainly talking about core audiences and then the outliers right? We don't want to feel like we are excluding people from our content, so show me the business case for who is watching a WebM video, is it someone on Firefox, who hasn't installed Flash ever on a desktop, that's a rare use case, right. I am sure it exists, but if you look -- and this is what I tell my stakeholders in any of the consultations that I do is to look at your analytics. What are your analytics telling you? I mean, we are lucky that this conversation around HTML5 today is 10 years from when all this first started right? We've had a decade of mistakes and most people, I would be surprised if they don't have any analytic data, right? If you don't have any analytics to know who is visiting your site, well let's start with that first; let's start capturing data of who is just coming to my site in general and so then you can start to make informed decisions. Like most stakeholders, when it comes to video content can easily find out who their audience is even if they are not capturing that data, now they can start to, within a month even, they would probably have enough data to know how to properly build a video solution that would cater to people going to their offering.
Troy Dreier: For those government clients you mentioned, isn't HTML5 the fallback?
Robert Reinhardt: I mean HTML5 and Flash, I mean which do you detect first and which do you position as the initial offering, it really depends on what you choose for your deployment. If you are going to -- I am a big fan of adaptive streaming, for example. So, if a client comes to me wanting adaptive streaming or we've worked out after discovering that adaptive streaming is what is going to best suit their audience, then you necessarily have a limited options there right? Adaptive streaming is only -- right now you have iOS with HLS or HTTP Live Streaming support and then you have Flash and Silverlight as browser plug-ins that can support it. Of those two plug-ins, most people are going to gravitate to Flash because of the ubiquity, so that's your choice right? If you want to invest in that as the delivery mechanism.
Troy Dreier: For HTML5 video, how many codecs are there?
Robert Reinhardt: For HTML5 video, I mean this is where from a W3 perspective, there is no codec recommendation, it's just a tag right and then they have attributes and events that they want -- enable a cross browser, so it is really up to the browser vendor to choose on the codec. So based on what browser vendors have selected, it's H.264 in varying profile support, so if I am on a mobile device, the profile support could vary from Baseline to Main to even High these days. I mean, a brand new tablet will playback High profile to WebM, which doesn't really have the same kind of profile schema that H.264 does. It does have some profile type of encoding parameters and then we have the oldest of the old, which is Theora, so there is essentially three families of codecs and any one of them can have potential variations. H.264, of course, being the one that has the most variations of the different profiles.
Troy Dreier: And of the codecs, which do you recommend to your clients?
Robert Reinhardt: If you don't have a huge content library, then encoding to multiple formats, I would actually recommend like have a fallback to WebM as a backup and you don't have any concerns about content protection. But when it comes to a general codec recommendation, I try to stay as H.264-centric as possible and I typically recommend having a set of Baseline encodes instead of Main profile encodes, so that you have Main for tablet and desktop and Baseline for all your smartphone devices.
Troy Dreier: Now, HTML5 video doesn't have pairity with Flash yet, what are some of the big features that it's missing and when do you think we'll have close enough to pairity for premium content owners?
Troy Dreier: I expected you to mention DRM in that answer.
Robert Reinhardt: DRM is definitely one of those things where I don't think we'll see HTML5 come to the table with a consistent uniform solution. By definition, it screams having something proprietary, that's why Flash and Silverlight have their own keys to working with encryption and right now with HTML5, we can use AES encryption to deliver a stream, but after it is encrypted, it is not bulletproof right? I mean it's not locked down like we could make sure that the subscriber is watching the stream, but once the content is coming into that device, we -- it really depends, I mean DRM, there's so many shades of grey there, right? I mean for me DRM is a lot about properly educating my clients and the affordability of the technology versus the loss of revenue because you don't have it in place.
Troy Dreier: So, the other hot topic at this conference is MPEG DASH, we are hearing a lot about it. I think there is still a lot of confusion about what it means, is MPEG DASH going to be a part of HTML5 video, is it going to replace HTML5 video? What is its future?
Robert Reinhardt: Let's just say Apple continues with HLS and that's all they do for adaptive streaming and they are like "Eh, MPEG DASH, we don't really need it, we've got our own solution already." But then we still have this environment where we have two streaming standards. Two is better than what we currently have, which is Adobe, Microsoft, and Apple all sort of pushing their own adaptive streaming manifest as Will -- I love the quote from Will Law at Akamai who's on the MPEG DASH panel here, he said that all the adaptive streaming specifications are 80% similar and 100% incompatible with each other. I love that because they all pretty much do the same thing right and they do it almost the same way with the video. They chunk up the video, most of them have rallied around H.264 and it's just a different specification for how those things are wrapped up, right?, but essentially the structure is the same. MPEG DASH would provide some kind of unified format that we could all use to publish, but it's codec agnostic, so if Google chooses to only have a MPEG DASH supported with WebM and Internet Explorer chooses to support MPEG DASH with H.264, I am still a bit -- what's the difference at that point? I still have to now have my content in two different flavors and so the adoption of MPEG DASH has a lot of requirements in order to make it work across the platforms. And the other issue is whether or not it would be backwards compatible. Would Apple only make it available in iPhone 6 and all the other iPhones that are already out there still have to use HLS? I mean it could be, regardless of what happens in MPEG DASH, it would be a multiyear development I think to see it become as predominant as any of the existing solutions. I mean we've had HLS already for a few years now and people already are adjusting their workflows to that, so the package has to be pretty tight and I think it will take some time in order to get agreement across the vendors. It would be nice if everyone, for example, used MPEG DASH with H.264, then we could at least have, you know, we could use a technology like Wowza to take H.264 files and transmux it into any number of wrappers that it needs, whether it's HLS, MPEG DASH, again it could raise more problems with workflow than it could solve.
Troy Dreier: Very good, well, thank you for joining me Robert; this is Troy Dreier coming to you from the red carpet.
FFmpeg might be a free download, but there's still a cost that goes with it. Robert Reinhardt explains why this encoder isn't right for every company.
The latest draft of the high-efficiency video coding codec appears to point to early 2013 for the final specification.
VideoRx.com's Robert Reinhardt explains how he's cleaning up the messy state of online video delivery.
The two server competitors provide interoperable delivery of adaptive streams during the 2012 London Olympics.
A panel compared Flash, HTML5, and Silverlight, despite the surprising news coming out hours before.
HTML5 video might be getting all the attention, but video publishers who want to serve the widest possible audience should make it the format of last resort. A presentation from Streaming Media East 2011 attempts to burst HTML5 video's balloon.