A Primer: H.264 vs. VP8 Quality Shootout and What it Means

1749731173_9616916fceNote: All comments are my own and may not reflect those of my employer.  This is a living position so I retain my right to edit and note the edits below as the conversation evolves.

For years when I worked on Web video and audio technologies from Xing MPEG, to Progressive Networks and RealNetworks, Windows Media and beyond, Jan Ozer was one of the most unbiased and critical analysts of subjective video quality during his tenure at PC Magazine, and more recently at Streaming Media.com.  Jan just posted his comparison of H.264 to VP8 and his results should start to refine the conversation around the future of video formats as Adobe, Apple, Google, Microsoft and the MPEG LA  set their positions for the next decade.  But before we get into the details, a quick primer on Video codecs, their impact on the Web and beyond.

Why do Codecs Matter?

Let me start by saying no reasonably consumer should need to think about video codecs.  Video should just work and quality should be good to great depending on the screen.  But for companies whose business models depend on the economics of video, the stakes are very real as you consider cost of compression in time, hardware requirements, and licensing costs.  Nearly all video compression formats today use perceptual methods to compress video into a “lossy” form by throwing out information the human eye isn’t likely to notice – not too dissimilar to how MP3 or JPEG images work today.  These methods are governed today by patents.

The reason we can enjoy great HD video quality and streaming video over the Internet today is really due to three things:

  1. Improved methods for compressing video and audio
  2. Faster processors to crunch large volumes of math representing the video & audio
  3. General acceptance and adoption in the industry (a long way of saying the vaunted “Ecosystem” word).

But to a consumer, the real value of implied or implicit standards in video is: “Just make sure I can watch it where I want, on what I want.”  Put another way, “Make it good enough, make it work”.  That’s where things start to break down.

A Brief History of a Decade of Web Video
A decade ago, three video formats battled it out – MPEG, Windows Media, and Real Video for postage stamped video delivery.  Each was instrumental in establishing the underpinnings of delivering video over the Web.  In the past 5 years, the quick maturation of Web video disrupted the marketplace with Flash promoting VP6, a technology Adobe had licensed from a smaller company called On2. Other codecs were available for Flash, but VP6 offered a better quality and cost structures. The explosion in popularity of Web video sites such as YouTube benefited from the dual ability to create their own branded video players and experiences on top of Flash and Flash Video on basic Web servers.

Meanwhile, MPEG and other codec technologies were absorbed into the emergent H.264 video standard.  Adobe, Apple, Microsoft, and others announced support for H.264 which is managed by a patent pool and licensing group called the MPEG-LA, LLC of which many companies including my employer are a member.  MPEG LA assembles patents for most consumer electronics-based digital video in the market today including those used in DVD and Blu-Ray, satellite and cable TV and receive royalties for their work in lieu of creating and promoting their own formats.

But others continued to work on their own format. On2 continued to plug away on their video codecs with primary licensor Adobe benefiting until being acquired by Google last year for their VP8 video codec.  What is VP8 you might ask?  Google also announced intent to release VP8 as WebM – an open-source, royalty free alternative to H.264 for use in HTML5 – the next-generation standard for Web browsers.

So a decade later we have H.264 (MPEG) and WebM (VP8) vying to be the de-facto video standard for the next-generation Web browser standard and beyond. For professionals, the dimensions they will evaluate on continue to be the same: Quality, Cost, and Reach.

Quality: When is it “Good Enough”?

The key value proposition for a codec provider a decade ago was who could deliver smoother, bigger, more TV-like video over the Web.  Each company strived to show how their video was better than the competition at delivering VHS, DVD, and later HD quality at a fraction the size.  But when does the video quality become “Good Enough”?  One could argue we’re approaching these limits already.  As Jan Ozer found in his recent evaluation of H.264 and VP8:

“H.264 still offers better quality, but the difference wouldn’t be noticeable in most applications.”

Based on Jan’s first evaluation, it sounds like VP8 is “Good Enough” in terms of quality.  More studies will need to be done but on face-value, the key points of differentiation have already shifted away from video quality to other dimensions of cost and reach.

Cost: What is the definition of “Free”?

In the past two years, the industry has seen rapid adoption of H.264 as an HD-ready alternative for consumer electronics and web-based experiences.  Many articles have been written regarding the pros/cons of the H.264 licensing terms which I won’t rehash here.  What’s different is Google’s approach.  With WebM they look to provide a free,royalty free route for licensing WebM and offering it up as a part of HTML5.

Reach: The Three Waves of Adoption

Next, you have industry adoption/reach.  Generally speaking, video formats seen three waves of adoption:

  1. Client Software – PC and/or Mac, new video formats today are first tested and proven for encoding, playback, and distribution via software encode/decode.  This is why certain video formats today eat up so much CPU – they run in software only.
  2. Servers & Solutions – Again, an offshoot of software, but here we see enhancements such as ability to deliver live content as well as on-demand from a server-type solution.  Integration partnerships ramp, solution providers and integration specialists for industries from video distribution to advertising support and you start to see Content Delivery Networks (CDNs) adopt the format en masse.
  3. Hardware Adoption – The last step is burning the format into silicon and/or enabling special software at the hardware level to accelerate at each point in the value chain: Creation/encoding, Distribution, and Client Playback.  This is the point at which you see everything from mobile phones to set-tops able to reasonably play back a format. Five years ago, a newer PC
    laying H.264 video would have pegged the processor; today’s latest smartphones can play it without issue. This because the device includes dedicated circuitry to decode the video while being conscious of things such as battery life.

Each of these waves are increasingly essential for any provider to play in.  The latter represents maturity.  Google has announced new hardware partnerships for Google TV that will offer hardware accelerated support for WebM “later”.  H.264 is further along in its maturity and adoption curve.

Where the Players and Lining Up

If you look across the landscape, the top players here are Adobe, Microsoft, Apple, Google and MPEG LA – and each has expressed their opinion on the matter.  So where do they stand?

Adobe – has announced they will support VP8/WebM in an undated future release of Flash Player.

Microsoft – will support choice of formats in Internet Explorer 9 and Silverlight .  Dean Hachamovitch recently posted the official response:

“When it comes to video and HTML5, we’re all in. In its HTML5 support, IE9 will support playback of H.264 video as well as VP8 video when the user has installed a VP8 codec on Windows.”

The Silverlight team also recently affirmed their position in a recent blog post.

Apple – So where does this leave Apple?  Steve Jobs recently responded to a customer mail noting:

“All video codecs are covered by patents. A patent pool is being assembled to go after Theora and other “open source” codecs now. Unfortunately, just because something is open source, it doesn’t mean or guarantee that it doesn’t infringe on others patents. An open standard is different from being royalty free or open source.”

Google – pretty clear considering they’re behind VP8/WebM and use H.264 today.  Some in the industry ask will Google force all YouTube video to WebM, putting pressure on Apple and others to follow?

MPEG LA – MPEG LA is reportedly investigating VP8/WebM in the interest of building a patent pool.

For those of us in the industry, more interesting times ahead, but this script feels a little like a Bill Murray film where we’re the weather man. Will hardware vendors be fast or slow to adopt VP8?  Will industry professionals adopt one or both, or wait and see?

What do you think?  Feel free to post here or email me at sean at seanalexander.com (fixing the at).

Update (05/23/11 7:30pm) – PC Magazine’s David Murphy also has a good recap though I think he’s oversimplifying the number of profiles that would be used in real-world use.

Photo: Fight for your Mind, by just.Luc on Flickr via Creative Commons license.

2 responses to “A Primer: H.264 vs. VP8 Quality Shootout and What it Means”

  1. Good overview. I don’t understand much about video but you seemed to sum it up.

  2. I am more interested in the decoder cpu and memory usage differences right now. As we switch to H.264 and MP4 for video storage (I work at a tv station), I’m noticing many older systems cannot playback the video, or playback is choppy due to the higher processing requirements of the client than say FLV or WMV9.