HEVC: Are We There Yet?
HEVC/H.265 burst onto the scene in January 2013, when it was released for production by the two standards bodies that helped create it, the International Telecommunication Union (ITU) and Moving Picture Experts Group (MPEG). Now, about 18 months after its release, how usable is HEVC? Well, actually, not very. This article will look at HEVC’s status in three segments: delivery to computers and notebooks, to mobile, and to OTT, identifying the trends and value propositions affecting each market.
Let’s start with a brief overview. HEVC/H.265 is the successor codec to AVC/H.264, and it’s designed to deliver two key benefits: equivalent quality to H.264 at approximately 50 percent the data rate, and 4K and ultimately 8K video. Reports of actual HEVC quality have varied. In May 2014, Dave Ronca, manager of encoding technology at Netflix, reported that HEVC yielded about the same quality as x264, while taking up to 10 times longer to encode. Most encoding vendors report that HEVC delivers between 30%-50% file size savings at the same quality as H.264, a range confirmed by most of my tests.
Like H.264, HEVC is a patent-encumbered technology. A group sponsored by MPEG LA includes most intellectual property (IP) stakeholders, though several IP owners have not joined the group. MPEG LA released a summary of expected license terms in January 2014, which included a $0.20 per unit royalty for encoder/decoder products after 100,000 units, and a $25 million initial annual cap, though final terms have not been set.
HEVC Delivery to Desktops and Notebooks
Are we there yet? Nope, nowhere close. But VP9 is.
Though mobile and OTT delivery get most of the headlines, most streaming producers send the bulk of their streams to computers and notebooks. Most computers and notebooks have the CPU horsepower to play at least 720p HEVC streams. Many can play 1080p and battery life is not as great a concern as it is in mobile markets. Encoding tools from multiple vendors are ready to go. All this being the case, you would think that the potential bandwidth savings of between 30% and 50% would accelerate HEVC usage by high volume producers, but that hasn’t been the case, for several reasons.
First, unlike H.264, there is no widely available free player for HEVC. Adobe has announced that it will support HEVC in its Flash-based premium distribution platform, Primetime, with no specific availability date, but haven’t announced HEVC support in the free Flash Player. According to DivX LLC, more than 4.7 million users have downloaded the HEVC Plug-in for their free DivX Player, which isn’t sufficient for an ESPN or CNN to start streaming HEVC solely to these players. VideoLAN’s VLC media player includes HEVC playback, but again doesn’t have the penetration necessary for general-purpose streaming.
Given the potential bandwidth savings, I wondered if any large content producers were seeking to directly license HEVC playback from IP vendors like Ittiam and Fluendo, who typically license codec technology and associated players to mobile, set-top box, and similar device manufacturers. Both reported some preliminary discussions, but nothing imminent.
These conversations raised the question about how a third-party HEVC decoder might fit into a web publisher’s decoding workflow, which is primarily Flash-based at this point for high-volume producers of premium, DRM-protected content. I spoke to several technologists who deemed it unlikely that a streaming producer could shoehorn third-party HEVC playback into their existing Flash-based players. Instead, they’d likely have to create a separate application that viewers would first have to download, which is most producers don’t like. For this reason, it seemed that most premium producers seeking to leverage HEVC in the short term would explore the relatively painless, though potentially more expensive, Primetime path than strike out on their own.
What about for content shipped without DRM? Here’s where things get interesting. We’re at a huge inflection point in general-purpose streaming where, due to the now ubiquitous support for the Media Source Extensions (MSE), HTML5 playback has become sufficiently robust for most producers of non-DRM content to stop using Flash. Specifically, MSE delivers both adaptive streaming and live, both with H.264 and VP9, an HEVC competitor that’s freely available in Google Chrome, Mozilla Firefox, and the Opera Browser.
YouTube has deployed VP9, and claims that it delivers H.264 quality at half the data rate and has cut rebuffer rates by 25% in developed markets, and 100% in emerging markets. One video technologist at another large publisher of non-DRM protected content, who asked not to be identified, says his company was also considering VP9.
YouTube claims that VP9 delivers the same quality at up to 50% compression gains.
Since VP9 is HTML5 compatible in the browsers that support it, it fits neatly into the new MSE framework. The obvious problem is lack of support in Internet Explorer and Safari, which YouTube addresses by supplying H.264 to those browsers, with Flash fallback to legacy browsers. Nonetheless, if I was transitioning to the Media Source Extensions, and my bandwidth costs were significant, I would think long and hard about adding an adaptive group of VP9 encodes, though not everyone agreed.
For example, an employee of one distributor of non-DRM movies (who also requested anonymity) says the company is currently evaluating HEVC, but didn’t expect to deploy to consumers until hardware decoders hit a critical mass. Though the company was actively exploring the transition to HTML5, which it hopes to accomplish in 2015, it is not evaluating VP9 and didn’t expect to deploy it.
A TV broadcaster with a huge web presence felt that “HEVC is inevitable -- not just for UltraHD, but across the entire spectrum of resolutions and quality levels,” says another anonymous source. “But I don’t think there will be a ton of movement until hardware decoders are ubiquitous in the consumer devices. We’re keeping our eye on it, but don’t anticipate doing much with it quite yet.” As for VP9? “In general VP9 (and VP8) is just not on our radar, and I don’t expect it to be anything we get involved with.”
What was the sentiment with major service providers? I checked with Casey Wilms, product manager of the media backbone group at Brightcove. Wilms reported that while Brightcove has received frequent inquiries about HEVC, few if any customers are asking about VP9. That said, Brightcove is very bullish on the Media Source Extensions and Wilms predicts that “we’ll be shocked at how fast” it takes off. While Wilms expects HEVC to dominate in broadcast markets, it’s not yet available for HTML5, and Brightcove will deliver whichever codec its customers require.
I spoke via email with Kaltura director of product management Michael Dale. He reported that Kaltura is also bullish on MSE, though Kaltura foresees “Flash being around as long as IE 10 or less is being used.” While Kaltura also sees limited demand for VP9, Dale acknowledged that “VP9 has a clearer upgrade path than HEVC, as it has a much smaller ecosystem and is less bounded against particular pieces of hardware.” Dale feels that VP9’s success will depend upon how effectively Google can leverage its part in the video ecosystem to promote the codec.
Finally, Tim Napoleon, chief strategist for digital broadcasting solutions provider All Digital, sees very little demand for VP9. “The immediate demand for H.265 is in high-resolution screens in the connected TV market,” Napoleon says. “Bandwidth savings for low resolutions for mobile application is not being widely asked for yet. H.264 is still widely supported and works well.”
Clearly, while VP9 is royalty free, it’s not completely free. There will be a significant encoding cost, and adding an adaptive group increases storage and administration costs, and decreases overall content cachability, potentially increasing delivery costs. There are also lingering patent violation claims from Nokia. Still, the math is pretty simple; if you’re moving to MSE and you could save significant dollars by cutting bandwidth on video delivered to Chrome, Firefox, and Opera, VP9 is worth a look. If you’re sticking with Flash, it’s not. Going forward, if Microsoft and Apple license HEVC for IE and Safari, which is widely expected, you’ve got more math to do, though it’s much more likely that your HEVC encodes will be able to perform double or triple duty for mobile and OTT.
Why would set-top box makers bother supporting UHD video when home bandwidth connections aren't nearly robust enough to carry it?
It's time to get started. Here's a look at best practices for encoding to H.265/HEVC with the MainConcept and x265 codecs
H.265 is coming to the Android operating system, as well, but it's nowhere to be seen on desktop or notebook computers.
HEVC/H.265 produces much smaller files, but there's a catch. If you want to get those results your encoding times will be much, much longer.
Whether they're talking about the cloud, 3D, 4K, or HEVC, IBC vendors and visitors look for the next shiny button to push. But will that advance the industry?
At this year's CES, some TV vendors asserted that HEVC was done for and VP9 would win out. What's happened with VP9 since then?
There's no reason for Adobe to add HEVC support to Flash, and that keeps HEVC from becoming relevant for general purpose streaming.
Which codec delivers better image quality? Which is more compatible? And what about Daala, the spoiler codec currently being developed from scratch?