Patented means proprietary, please to not try to make words mean things that they do not.
No, patent means patented. There's a qualitative difference between a format for which there are public and publshed interoperable standard, and one where the implementation details are private or only avialable under a specific license. You may not care for either model, but it's certainly a meaningful distinction, and a longstanding one in the digital media world. When people in the vieo industry speak of an "open standard" they mean publically available specifications and patent licenses available under RAND terms. A propritary codec would be used to describe, say, RealVideo 10 or Apple's ProRes, for which there isn't bitstream documentation or RAND licensing availble.
With an open standard, everyone's on equal ground in building interoperable implementations without any reverse engineering, and has equal abilty and pricing for licensing the patents.
The issue of whether those patents have a fee or not is obviously important, but somewhat orthogonal to openness. One could certainly have a free-to-implement technology that isn't documented, and hence wouldn't be considered "open."
And Theora is certainly patented as well; On2 has released their patents under an extremely flexible license, but they're still valid.
I repeat my assertion that Ogg Theora is already good enough for me, and likely is good enough for many besides myself, who do not care much about 3 DB more or less of streaming bandwidth, and who do care about freedom from proprietary restrictions and patent fees for video codecs.
I've never heard streaming bandwidth described in dB. Interesting metric; so 3 dB would be ~2x bandwidth difference at the same quality? Kind of elegent; I normally talk about that in terms of percentage, but since improvements are measured in dB, it could apply either way.
FWIW, codec engineers sweat blood for a 0.1 dB improvement. The cable industry has spent multiple billions of dollars to upgrade to H.264 set top boxes and infrastructure to get that ~3-4 dB improvement of H.264 over MPEG-2, expecting a much bigger payoff due to additioanl channels/services they can sell with those savings.
Anyway, if Theora does what you want it do, use it with my blessing. Good enough is by definition good enough. Like I said earlier, I work on Silverlight, and we've already got the infrastructure in Silverlight for 3rd parties to add new codec and format support in managed code.
My concern is mainly that a lot of people seem to be thinking that Theora is capable of things it isn't and won't be capable of. To whit: Theora isn't ever going to be competitive with H.264 High Profile in compression efficiency. While it's certainly capable of futher improvement, H.264 implementatiosn are improving rapidly as well, so I doubt it'd ever need less than 2x the bandwidth for a particular quality level compared to best H.264 implementations at the time.
For the business models I've run some quikc numbers on, the extra bandwidth cost of Theora would cost more than any H.264 license fees saved.
Thus mainstream media sites, like YouTube, don't have any business reasons to adopt Thera; it'd be a net negative on their profitability.
If you think I'm mistaken on any of the above, I'd be very interested in disucussing your perspective. If you're asserting that there are markets where the above factors don't matter much, then I agree with you.
But if it's really important for this community to have a competitive codec without patent licensing requirements, then Theora (at least a Theora 1.0 bistream compatible version) may be a distraction.
I don't have a lot of hope for Dirac either; I've not seen any indication of a new approach to the challenges of marrying wavelts with motion estimation; once your intra and inter block sizes are radically different, things get quite challenging. Theora is likely to remain a superior choice than Dirac.
Yeah, they might, but that doesn't mean the WMDRM scheme would be workable for Bluray discs themselves.
Nope, that's AACS. Microsoft is a participant in the group that made it, however. I beleive WMDRM is an approved DRM for transcoding from AACS, which should matter when AACS Managed Copy finally becomes avaialble.
Again, so what? People would have ripped the analog signal, encoded it, and we'd be in the same boat. They can try to make things more painful for consumers, which is what DRM already does, but that will only continue to hurt their business. People will go on finding ways to do what they want anyway, they'll avoid legitimate products more the more painful they are, and their business model will fall apart.
Yep. It's a corundrum. I'm glad not to be in that industry myself. My general advice to them has been to focus on having a higher total value than pirated content, including higher quality, better features, extras, etcetera. In order to reduce piracy, the pain of DRM should be less than the value-add of commerical content. A lot of that gets down to business and usage models more than technology, as Hulu's success has shown.
Commerical services win when they make the Pirate Bay and Bittorrent seem like lesser versions of the content and generally too much hassle.
The counter-example is of course Audio CD protection schemes, which made the actual purchased CD much less valuable than downloaded versions, and added much more incentive to piracy than barriers against it.
This is totally irrelevant. HTML5 is a new standard and can use whatever is best for the job, not what was popular for whatever reason before. Because otherwise we should use the crappy Sorenson H.263 (old flash codec), it's probably still the #1 codec on the web.
Not if you care about playback on devices like phones, media players, and Netbooks which really count on having hardware decode for media playback.
The fact is that for something so important and so widely used as the web, it is indispensable that the standard can be freely used by everyone, and not controlled by whoever happens to have the patents for some video format.
Well, I don't think H.264 or any MPEG-LA codec can be said to be "controlled" - it's RAND licensing and available to anyone. It's not free to implement, but it's not propritary either. Lots of ISO standards are like that.
Freedom of the web is much much much more important than picture quality. Let alone that if ogg is used in HTML5, it will attract a lot of research and very quickly we'll have a high quality, free to use video format.
Well, there's plenty of companies who care a whole lot about quality and delivery costs. Theora is certainly capable of high quality; it can do that already with sufficient bitrates.
The real challenge is in compression efficiency, and that's fundamentally constrained by the bistream syntax. Optimizations can converge on the theoretical limits of a codec, but those are going to be a lot lower than H.264, and lower yet than H.265 in a few more years. And while Theora may get a lot of attention and tuning, H.264 is already getting LOTS of that from multiple vendors and groups competing to build the best implementations. H.264 is getting better at least as quickly as Theora is.
But certainly, Theora is already more than good enough for plenty of tasks. Wikipedia is committed to using it, and for short clips of smaller frame size, it'll be fine. It's going to be better than MS MPEG-4v1 and the original RealVideo codec that I had to use at the birth of web video, and there's much more bandwidth available as well.
But H.264 High Profile will be able to deliver equivalent quality to Theora at a half to a third the bitrate, and broader compatibilty with existing devices. How important those considerations are will vary by market, but are very important to some big markets. YouTube and Hulu certainly aren't going to double their bandwidth budget and end-user connection speed requirements.
Btw, your codec list is misleading. Flash does support H.264 but still its old format (H.263) is more widely used. Moreover it supports HTTP and 99% of the video sites (including youtube) use HTTP to serve videos. RTMP is used for "real" streaming with Flash Media Server (Adobe's streaming server that few people use due to high price) and RTMPe that you mention, is the encrypted version of RTMP, used by very very few.
Now that YouTube is using H.264, I'm sure that the eyeball-hours of H.264-in-Flash are a lot higher than H.263-in-Flash today. VP6 versus H.264 is the more interesting competition. YouTube hasn't used it, but most of the big Flash media sites like Hulu skipped H.263 entirely and started with VP6.
Alter Relationship on Friday July 03, @02:23AM (#28569633) Homepage Well, we can say even Microsoft admitted H264 and MPEG4 base is there to stay no matter what they do so they declared "fallback". It is not just Microsoft, Real Networks decided on h264/AAC long time ago. The high bandwidth is aac+ and h264 based codec.
It's not that MPEG-4 is just a fallback. The fragmented MPEG-4 file format is used for Smooth Streaming, which is the premimum media experience for Silverlight.
ASF was a good design for a "bit pump" to deliver server scalability a decade ago, but fMP4 is much better to enable byte-range addressable self-contained fragments.
TV and Movie companies already paid huge sums of money to professional licensing of mpeg. In fact, you have paid too with each device you own which plays mpeg4.
H.264 licensing really isn't that expensive either. It's a small portion of the cost of the technology required for digital distribution, and the competitive effects of being an interoperable standard certainly reduced prices much more than the cost of the license fee.
H264 was essentially chosen because it is not tied to single vendor, documented and there to stay.
...and has industry leading bitrate efficiency. That's a huge part of it.
1) VC-1 is used on Blu-ray 2) The chair of the H.264 committee has been a Microsoft employee for the last decade 3) H.264 is supported in Xbox, Zune, Win 7, MediaRoom, etcetera. 4) H.264 was announced for Silverlight before Silverlight 2 shipped. It just took a while to get implemented.
You sure do put a lot of energy into slagging Ogg, and you consistently neglect mention the advantage Ogg has over H.264: it is unencumbered by patents and therefore free for anybody to encode and/or play, on any hardware they wish.
Oh, I don't have anything against Theora per se, nor Ogg in general. It's just people keep having highly unrealistic hopes for what it can do in terms of compression efficiency and ecossytem.
Codecs are hard, and it does no one any benefit to assume they're capable of things they simply aren't.
The Xiph blog posts on their optimization process for Theora have been excellent reading, and they've done really good work. But the bitstream itself simply isn't capable of what modern codecs are capable of already. I'm sure Theroa will continue to improve, but I don't see any reaosn why H.264 won't see improvements at least as quickly.
And H.265 is already in development, targeting new bistream features that will add further substantial efficiency improvements.
I for one, am perfectly happy to burn a little extra bandwidth for that, and anyway I not buy your assertion that Ogg cannot close the bandwidth gap over time..
After all, you are a Microsoftie with a vested interest in keeping video proprietary
Eh, I work on Silverlight, where we have the Raw AV pipeline for managed code decoders. It's be trivial for any customer add Theora support to Silverlight if they want it. If anything, Theora would be a competitive advantage for Silverlight.
Also, I don't think anyone is talking about propritary codecs here, except for perhaps VP6. VC-1 and H.264 are both international standards, with licensing handled by MPEG-LA. They are patent encumbered, but are not propritary any more than MP3 or ASP are.
Both Apple and AACS are reliant on support from platforms that Microsoft can't get their hands on. You wouldn't have FairPlay without Macs/iPods, and you wouldn't have AACS without stand-alone bluray players.
Actually plenty of Blu-ray and other devices support WMDRM via DLNA or Media Extender protocols. This stuff gets deployed pretty broadly on devices you may not have ever thought of as having DRM, like phones. Most high-end Nokia phones have both WMDRM and OMA.
I'm kind of inclined to say "so what?" I mean, I understand what you're saying, and you're not wrong there, but that doesn't mean DRM was generally a positive development. As far as I'm concerned, its main effect has been to delay the development of rational business models by preserving the illusory view that media companies have that they're in a position to set their own terms.
Well, the studios could also have said "analog only!" and designed technologies that don't interoperate well with the PC ecosystem. Have you ever read the Digital Cinema Initative specs? 4096x2048 12-bit JPEG2000 using the XYZ colorspace? It's chock full of nerdy goodess, but it also reads like exactly what a studo would design if trying to make a technolgoy as painful for home use as possible.
Asking if DRM was good or bad on the whole is like talking about whether we're better off today because of Napolean:). Without a Napolean-free or a DRM-free world to compare it's hard to say what the alternatives would look like.
Why would anyone put serious effort into competing with them on Windows?
Apple did. AACS (Blu-ray) did. Whether it makes sense in theory, folks have done it in practice:). Anyway, Windows output protection etcetera APIs are documented anyway.
I'm sorry, but I don't see how you can even claim that in good faith. The bits can be liberated just fine without DRM. There's no way to argue that DRM isn't about trying to control distribution and use.
I don't think we're disagreeing here. Hollywood wasn't inclined to allow content to play in PCs at all unless they were comfortable with available security. They could have refused any CSS licenses for anything other than a hardened CE player with analog outputs. I don't know if that would have stopped piracy much, but it certainly would have made a legitimate market for digital content much slower to develop.
...but insofar as there's no alternative, it's because Microsoft has a lock on the market. You can't really have a workable open source DRM scheme, and Apple has their own. So anyone trying to compete with Microsoft to be the DRM provider has to compete against Microsoft on their own platform.
Well, it's not like Microsoft was the only company working on DRM or with the potential to do so. Apple's had a lot of traction, but have decided not to allow any other company to publish to Fairplay. There's OMA as an open platform, and Sony's attempt with OpenMG. Adobe's DRM systems in Flash and (different one) AIR. And all of those worked and continue to work fine on Windows, of course. I'd say it's really been Apple and Adobe, and to a lesser degree Sony, really trying to build a walled garden. And WMDRM has always had RAND licensing; Apple could have licensed it and implemented it themselves if they were so inclined.
DRM was more about liberating the bits from the discs and the fixed-function CE devices. The real competition for WMDRM was DVD and CSS; content provdiers had to feel that their content was going ot be reasoanble secure in order to let it on anything Turing Complete device. The studios giving up on DRM isn't a loss for anyone, I don't think, but a win for anyone making file-based playback devices.
I don't know if it's possible to make significant, sustained revenue from DRM. It's one of those ecosystem components that unblocks a lot of other stuff, but doesn't appear to have have financial value it itself.
I doubt they paid that much. Like On2 with VP6, Sorenson's model was to license the decoder and a consumer grade encoder on the cheap, and then make the real money by selling professional grade encoders once there was a good installed base of players.
Since I said CPU, I thought I was clear that I meant decoding in software.
Sure. But no device meant for media playback is going to use software decoding. It'll already have a decoder ASIC, typically with at least H.264 Baseline and VC-1 Main Profile. Either will offer better quality and much longer battery life in an ASIC implementation compared to a software Theora decoder.
Also, about the battery life -- I can get about 5-6 hours of playback time decoding purely on the CPU with an Atom N280.
IS that with a GMA500? That's capable of hardware accelerated MPEG-2, VC-1, and H.264 decoding.
What are your mobile devices, and what's your media player? All the current ones that are meant for media playback include H.264 ASICs. And those are getting crazy good; the Zune HD's going to support 720p HD playback using the NVidia Tegra.
Lacking an ASIC, any Theora on devices would need to be done in sofware, and even a simple codec can be extremely taxing on a 400-600 MHz ARM. Even if it's playable it's going to eat battery like no tomorrow. I can imagine a really good implementation being able to maybe do 320x240 30p 500 Kbps Theora in software on a 600 MHz ARM, but that'd require a whole lot of tuning.
The VP3 bitstream predates device media, so I doubt it has any particular design tuning for them. It's very much a codec designed for x86, with a PPC port.
The low bitrate, for whatever reason is keeping to the specs they've been using since launch, which are using the xvid implementation of old Sorenson Spark H.263 v1/MPEG-4 Part 2 Short Header. Maybe for device compatibility? Anyway, That's a codec about as old as the Theora bitstream, so we wouldn't expect it to be much better.
But I don't know that YouTube thinks it's "good enough" - they're offering higher quality modes, and that's what you get by default on the iPhone and other platforms. For whatever reason they're keeping around a legacy version, likely backwards comaptibility with some clients that don't do H.264 for whatever reason.
For the their high quality streams, Theora isn't competitive in quality. And for the highly compatible streams, Theora isn't competitive in compatibility.
So YouTube saying that Theora doesn't make sense for them makes sense to me. Therora doesn't an advantage in quality or compatibility for the streams they're doing.
Also, Big Buck Bunny isn't the best clip to extrapolate from, as it's really high quality lossless animation. To really see what YouTube needs to handle, try some lousy webcam, DV, and VOB rips. That's where H.264's in-loop deblocking filter give it a big advantage over other codecs, because it just gets smoother intead of blocky as the content gets more challenging.
Not to dismiss the excellent development work Xiph has done on Theora. The posts have been a fascinating read. But it's not plausible to me that anyone can make a business case for Theora over H.264, VC-1, or ASP licensing is available; the reduced bandwidth costs would be bigger than the actual real-world licensing fees for the real world examples I've thought of.
Theora's sweet spot would be in cases where MPEG-LA codec licenses simply aren't available for whatever reason. I imagine a fully refined Theora decoder would need fewer MIPS/pixel than H.264 High Profile, and perhaps even Baseline. But even in those cases, VC-1 Main Profile will probably offer similar performance with significantly better efficiency.
Does Ogg even have hardware acceleration at this point?
Nope. I don't know if anyone's even scoped a hardware implementation of VP3. There have been some VP6/7 DSP implementations, but no ASIC ones (ASIC have better power consumption).
Now, Theora is a pretty simple codec, so doing it in hardware would be a lot simpler than H.264 and probably simpler than VC-1. But it can take quite a few engineering years to refine a decoder for performant playback.
Of couree, performance isn't just the video decoder. It's the video and the audio decoder, and the whole pipeline to make sure you get smooth in-sync playback. Media pipelines are really hard, particularly if you're trying to implement them as part of a browser rendering model that never had to worry about timestamps, decoder buffers, etcetera.
Well they're not pushing it too hard anymore, but that's really because they already lost on the audio side. Their hopes for locking up online music sales died when the major labels agreed to sell without DRM. Video may not be all that far behind.
Once the labels don't require DRM to release content, than the PC and devices are a healthy market for commercial music. Mission accomplished.
Having a good DRM still makes other business models possible, like ZunePass, and it's not like there's some other braodly implemented standards-based DRM as an alternative.
Generally the stuff Windows Media gets used for exclusively is stuff where there's not a viable alternative.
It hurts Microsoft, but particularly because their formats are more open than the proprietary alternative Microsoft is pushing.
Microsoft really isn't "pushing" Windows Media that much anymore. Zune and Xbox already support MPEG-4 and H.264, as will Silverlight 3 and Windows 7.
Really, the money is in having a good flexible media platform much more than in owning one. Windows Media exists to solve problems that weren't otherwise being solved. But the chair of the committee that designed H.264 has been a Microsoft employee for the last decade.
The only company that really made a lot of money on media platform licensing has been On2 with VP6, and their high licensing costs (like multiple $K per core for enterprise encoding) were one big reason that Adobe adopted H.264.
Well, it's not like YouTube sells more ads with better looking video, and I doubt 90% of the uploads get watched more than a dozen times. They probably have some pretty deep metrics about the watts/cents per minute of video encoding and tune for that.
YouTube is also really only a good example of YouTube, since they're a massively money-losing operation funded by a very rich company. No one else does it like YouTube, and ever other video site is going to average a lot higher views/clip, so they can afford more CPU time to improve quality.
Or maybe they're just not very good at video compression:).
Beyond B-frames, they're not using 8x8 blocks or CABAC entropy coding, both of which can offer substantial efficiency improvements.
In recent years, they have both been proponents of DRM at some point, both support their own proprietary formats (Microsoft with WMA/WMV/ASF
The WMV 9 video codec has been standardized as VC-1 via SMPTE. Its licensing is handled by MPEG-LA, like MPEG-2, MPEG-4 part 2, and H.264.
Apple with QuickTime, and AAC
QuickTime was the basis for and largely the same as the MPEG-4 file format.
AAC==Advanced Audio Codec. It's a MPEG technology originally created for MPEG-2. Apple is one of many implementors, and was not the primary developer (which was AT&T, I believe).
Microsoft won't support a codec that makes the web more reliable for non-Windows users - especially Linux users.
Actually, the currently in beta Silverlight 3 supports WMV and H.264, and has an extensible Raw AV pipeline that would make implementing the Ogg codecs inside of Silverlight trivial.
And that technology is already impleented in Moonlight, so Linux users have access to it as well.
So, Microsoft is actively exactly doing that, and includig Mac and Linux users.
Theora is a pretty weird codec. It's got a strange scan pattern, goes from bottom up instead of top down, and has lots of other rather unique aspects that could make a silicon implementation harder than its relative simplicity would suggest.
The funny thing is there are royalty free codecs out there.
I don't think anyone's ever asserted patents against MPEG-1, H.263, or MP2 audio, and any patents would be due to expeire soon anyway.
If being patent and license free is the paramount concern, there's plenty of choices with much more mature implementations than Theora. I don't understand the exclusive focus on just a few immature codecs like Theora and to a lesser degree Dirac.
Now, the bigger challenge is weighing the cost of patent licensing versus the cost of bandwidth. H.264 pencils out cheaper than Theora in practice when bandwidth costs and audience size are compared (lower bandwisth==bigger audience with sufficient bandwidth for the content).
Theora is based on VP3, which was both patented and commercially distributed for a number of years. If VP3 had been infringing someone else's patents, then they would likely have sued back when a company was making money from it.
Well, I don't believe On2 ever made any money from VP3. VP6 with Flash 8+ was their big moneymaker. They also licensed the TrueMotion codecs for games back in the day, and even licensed their TrueMotion RT codec fof $5M to Microsoft for their never-shipped NetShow Theater product.
But it was pretty much the DotCom investment bubble that floated them from the late 90's through Flash 8. And they shrunk by a good 90% in staff from 2000 until VP6.
I met with them at a trade show a few months ago and realized I'd actually been having business dealings with On2 (to back when they were called the Duck Corporation) for longer than any current employee has been there. It's weird telling a company their own war stories back to them:).
By Windows Media do you mean DirectShow? DirectShow is extensible to 3rd party codces, and there are Theora implementations. But not for Windows Media.
There are really only two significant video formats today for web streaming: Mpeg4/H.264 with MP3 or AAC audio is technically superior; Ogg/Theora with Vorbis audio is freer.
Ogg use on the internet is a rounding error at best; RealMedia still gets more use (very popular in internet cafes in China for some reason).
The three primary media formats/codecs are MPEG-4 + H.264 (QuickTime, plus Flash and soon Silverlight 3 via progressive download), RTMPe + H.264 (Flash uses MPEG-4 files but a propritary protocol), and Windows Media + VC-1. Move Networks + VP7 (ala ABC.com) also pulls in million of eyeball/hours a month, certainly more than Ogg at this point.
I'd say Ogg is #5 at best today. #6 if you count torrents and hence MPEG-4 part 2.
As for Microsoft support, that's becoming pretty codec neutral. Silverlight 3 (currently in beta) supports both H.264 and has a Raw AV pipeline allowing arbitrary codecs in managed code to be added to any Silverlight player. So adding Theora/Vorbis or any other codec, format, and protocol can be done inside the Silverlight sandbox by any third party.
Ignoring the tremendous improvements in the Thusnelda branch, if YouTube suddenly switched from severe H.26whatever overcompression to stock Theora with optimal settings (and everyone had libtheora and HTML 5 browsers), no one would notice the difference.
Untrue. Xiph has made heroic progress with Theora, but it's still a decade-old codec design and bitstream, and it's hard to imagine it catching up with xvid, let alone a good H.264 implementation.
YouTube certainly has quality issues, but things can be bad in more than one way at a time. There's nothing that less efficient codec would help them with. Note their top bitrate is 1280x720p30 at 2 Mbps.
Some samples compared Xiph's latest demo clips, with the same source encoded with VC-1 and x264 are here:
Slashdot Mirror
Patented means proprietary, please to not try to make words mean things that they do not.
No, patent means patented. There's a qualitative difference between a format for which there are public and publshed interoperable standard, and one where the implementation details are private or only avialable under a specific license. You may not care for either model, but it's certainly a meaningful distinction, and a longstanding one in the digital media world. When people in the vieo industry speak of an "open standard" they mean publically available specifications and patent licenses available under RAND terms. A propritary codec would be used to describe, say, RealVideo 10 or Apple's ProRes, for which there isn't bitstream documentation or RAND licensing availble.
With an open standard, everyone's on equal ground in building interoperable implementations without any reverse engineering, and has equal abilty and pricing for licensing the patents.
The issue of whether those patents have a fee or not is obviously important, but somewhat orthogonal to openness. One could certainly have a free-to-implement technology that isn't documented, and hence wouldn't be considered "open."
And Theora is certainly patented as well; On2 has released their patents under an extremely flexible license, but they're still valid.
I repeat my assertion that Ogg Theora is already good enough for me, and likely is good enough for many besides myself, who do not care much about 3 DB more or less of streaming bandwidth, and who do care about freedom from proprietary restrictions and patent fees for video codecs.
I've never heard streaming bandwidth described in dB. Interesting metric; so 3 dB would be ~2x bandwidth difference at the same quality? Kind of elegent; I normally talk about that in terms of percentage, but since improvements are measured in dB, it could apply either way.
FWIW, codec engineers sweat blood for a 0.1 dB improvement. The cable industry has spent multiple billions of dollars to upgrade to H.264 set top boxes and infrastructure to get that ~3-4 dB improvement of H.264 over MPEG-2, expecting a much bigger payoff due to additioanl channels/services they can sell with those savings.
Anyway, if Theora does what you want it do, use it with my blessing. Good enough is by definition good enough. Like I said earlier, I work on Silverlight, and we've already got the infrastructure in Silverlight for 3rd parties to add new codec and format support in managed code.
My concern is mainly that a lot of people seem to be thinking that Theora is capable of things it isn't and won't be capable of. To whit:
Theora isn't ever going to be competitive with H.264 High Profile in compression efficiency. While it's certainly capable of futher improvement, H.264 implementatiosn are improving rapidly as well, so I doubt it'd ever need less than 2x the bandwidth for a particular quality level compared to best H.264 implementations at the time.
For the business models I've run some quikc numbers on, the extra bandwidth cost of Theora would cost more than any H.264 license fees saved.
Thus mainstream media sites, like YouTube, don't have any business reasons to adopt Thera; it'd be a net negative on their profitability.
If you think I'm mistaken on any of the above, I'd be very interested in disucussing your perspective. If you're asserting that there are markets where the above factors don't matter much, then I agree with you.
But if it's really important for this community to have a competitive codec without patent licensing requirements, then Theora (at least a Theora 1.0 bistream compatible version) may be a distraction.
I don't have a lot of hope for Dirac either; I've not seen any indication of a new approach to the challenges of marrying wavelts with motion estimation; once your intra and inter block sizes are radically different, things get quite challenging. Theora is likely to remain a superior choice than Dirac.
Yeah, they might, but that doesn't mean the WMDRM scheme would be workable for Bluray discs themselves.
Nope, that's AACS. Microsoft is a participant in the group that made it, however. I beleive WMDRM is an approved DRM for transcoding from AACS, which should matter when AACS Managed Copy finally becomes avaialble.
Again, so what? People would have ripped the analog signal, encoded it, and we'd be in the same boat. They can try to make things more painful for consumers, which is what DRM already does, but that will only continue to hurt their business. People will go on finding ways to do what they want anyway, they'll avoid legitimate products more the more painful they are, and their business model will fall apart.
Yep. It's a corundrum. I'm glad not to be in that industry myself. My general advice to them has been to focus on having a higher total value than pirated content, including higher quality, better features, extras, etcetera. In order to reduce piracy, the pain of DRM should be less than the value-add of commerical content. A lot of that gets down to business and usage models more than technology, as Hulu's success has shown.
Commerical services win when they make the Pirate Bay and Bittorrent seem like lesser versions of the content and generally too much hassle.
The counter-example is of course Audio CD protection schemes, which made the actual purchased CD much less valuable than downloaded versions, and added much more incentive to piracy than barriers against it.
This is totally irrelevant. HTML5 is a new standard and can use whatever is best for the job, not what was popular for whatever reason before. Because otherwise we should use the crappy Sorenson H.263 (old flash codec), it's probably still the #1 codec on the web.
Not if you care about playback on devices like phones, media players, and Netbooks which really count on having hardware decode for media playback.
The fact is that for something so important and so widely used as the web, it is indispensable that the standard can be freely used by everyone, and not controlled by whoever happens to have the patents for some video format.
Well, I don't think H.264 or any MPEG-LA codec can be said to be "controlled" - it's RAND licensing and available to anyone. It's not free to implement, but it's not propritary either. Lots of ISO standards are like that.
Freedom of the web is much much much more important than picture quality. Let alone that if ogg is used in HTML5, it will attract a lot of research and very quickly we'll have a high quality, free to use video format.
Well, there's plenty of companies who care a whole lot about quality and delivery costs. Theora is certainly capable of high quality; it can do that already with sufficient bitrates.
The real challenge is in compression efficiency, and that's fundamentally constrained by the bistream syntax. Optimizations can converge on the theoretical limits of a codec, but those are going to be a lot lower than H.264, and lower yet than H.265 in a few more years. And while Theora may get a lot of attention and tuning, H.264 is already getting LOTS of that from multiple vendors and groups competing to build the best implementations. H.264 is getting better at least as quickly as Theora is.
But certainly, Theora is already more than good enough for plenty of tasks. Wikipedia is committed to using it, and for short clips of smaller frame size, it'll be fine. It's going to be better than MS MPEG-4v1 and the original RealVideo codec that I had to use at the birth of web video, and there's much more bandwidth available as well.
But H.264 High Profile will be able to deliver equivalent quality to Theora at a half to a third the bitrate, and broader compatibilty with existing devices. How important those considerations are will vary by market, but are very important to some big markets. YouTube and Hulu certainly aren't going to double their bandwidth budget and end-user connection speed requirements.
Btw, your codec list is misleading. Flash does support H.264 but still its old format (H.263) is more widely used. Moreover it supports HTTP and 99% of the video sites (including youtube) use HTTP to serve videos. RTMP is used for "real" streaming with Flash Media Server (Adobe's streaming server that few people use due to high price) and RTMPe that you mention, is the encrypted version of RTMP, used by very very few.
Now that YouTube is using H.264, I'm sure that the eyeball-hours of H.264-in-Flash are a lot higher than H.263-in-Flash today. VP6 versus H.264 is the more interesting competition. YouTube hasn't used it, but most of the big Flash media sites like Hulu skipped H.263 entirely and started with VP6.
Alter Relationship on Friday July 03, @02:23AM (#28569633) Homepage
Well, we can say even Microsoft admitted H264 and MPEG4 base is there to stay no matter what they do so they declared "fallback". It is not just Microsoft, Real Networks decided on h264/AAC long time ago. The high bandwidth is aac+ and h264 based codec.
It's not that MPEG-4 is just a fallback. The fragmented MPEG-4 file format is used for Smooth Streaming, which is the premimum media experience for Silverlight.
ASF was a good design for a "bit pump" to deliver server scalability a decade ago, but fMP4 is much better to enable byte-range addressable self-contained fragments.
TV and Movie companies already paid huge sums of money to professional licensing of mpeg. In fact, you have paid too with each device you own which plays mpeg4.
H.264 licensing really isn't that expensive either. It's a small portion of the cost of the technology required for digital distribution, and the competitive effects of being an interoperable standard certainly reduced prices much more than the cost of the license fee.
H264 was essentially chosen because it is not tied to single vendor, documented and there to stay.
...and has industry leading bitrate efficiency. That's a huge part of it.
1) VC-1 is used on Blu-ray
2) The chair of the H.264 committee has been a Microsoft employee for the last decade
3) H.264 is supported in Xbox, Zune, Win 7, MediaRoom, etcetera.
4) H.264 was announced for Silverlight before Silverlight 2 shipped. It just took a while to get implemented.
You sure do put a lot of energy into slagging Ogg, and you consistently neglect mention the advantage Ogg has over H.264: it is unencumbered by patents and therefore free for anybody to encode and/or play, on any hardware they wish.
Oh, I don't have anything against Theora per se, nor Ogg in general. It's just people keep having highly unrealistic hopes for what it can do in terms of compression efficiency and ecossytem.
Codecs are hard, and it does no one any benefit to assume they're capable of things they simply aren't.
The Xiph blog posts on their optimization process for Theora have been excellent reading, and they've done really good work. But the bitstream itself simply isn't capable of what modern codecs are capable of already. I'm sure Theroa will continue to improve, but I don't see any reaosn why H.264 won't see improvements at least as quickly.
And H.265 is already in development, targeting new bistream features that will add further substantial efficiency improvements.
I for one, am perfectly happy to burn a little extra bandwidth for that, and anyway I not buy your assertion that Ogg cannot close the bandwidth gap over time. .
After all, you are a Microsoftie with a vested interest in keeping video proprietary
Eh, I work on Silverlight, where we have the Raw AV pipeline for managed code decoders. It's be trivial for any customer add Theora support to Silverlight if they want it. If anything, Theora would be a competitive advantage for Silverlight.
Also, I don't think anyone is talking about propritary codecs here, except for perhaps VP6. VC-1 and H.264 are both international standards, with licensing handled by MPEG-LA. They are patent encumbered, but are not propritary any more than MP3 or ASP are.
Both Apple and AACS are reliant on support from platforms that Microsoft can't get their hands on. You wouldn't have FairPlay without Macs/iPods, and you wouldn't have AACS without stand-alone bluray players.
Actually plenty of Blu-ray and other devices support WMDRM via DLNA or Media Extender protocols. This stuff gets deployed pretty broadly on devices you may not have ever thought of as having DRM, like phones. Most high-end Nokia phones have both WMDRM and OMA.
I'm kind of inclined to say "so what?" I mean, I understand what you're saying, and you're not wrong there, but that doesn't mean DRM was generally a positive development. As far as I'm concerned, its main effect has been to delay the development of rational business models by preserving the illusory view that media companies have that they're in a position to set their own terms.
Well, the studios could also have said "analog only!" and designed technologies that don't interoperate well with the PC ecosystem. Have you ever read the Digital Cinema Initative specs? 4096x2048 12-bit JPEG2000 using the XYZ colorspace? It's chock full of nerdy goodess, but it also reads like exactly what a studo would design if trying to make a technolgoy as painful for home use as possible.
Asking if DRM was good or bad on the whole is like talking about whether we're better off today because of Napolean :). Without a Napolean-free or a DRM-free world to compare it's hard to say what the alternatives would look like.
Why would anyone put serious effort into competing with them on Windows?
Apple did. AACS (Blu-ray) did. Whether it makes sense in theory, folks have done it in practice :). Anyway, Windows output protection etcetera APIs are documented anyway.
I'm sorry, but I don't see how you can even claim that in good faith. The bits can be liberated just fine without DRM. There's no way to argue that DRM isn't about trying to control distribution and use.
I don't think we're disagreeing here. Hollywood wasn't inclined to allow content to play in PCs at all unless they were comfortable with available security. They could have refused any CSS licenses for anything other than a hardened CE player with analog outputs. I don't know if that would have stopped piracy much, but it certainly would have made a legitimate market for digital content much slower to develop.
...but insofar as there's no alternative, it's because Microsoft has a lock on the market. You can't really have a workable open source DRM scheme, and Apple has their own. So anyone trying to compete with Microsoft to be the DRM provider has to compete against Microsoft on their own platform.
Well, it's not like Microsoft was the only company working on DRM or with the potential to do so. Apple's had a lot of traction, but have decided not to allow any other company to publish to Fairplay. There's OMA as an open platform, and Sony's attempt with OpenMG. Adobe's DRM systems in Flash and (different one) AIR. And all of those worked and continue to work fine on Windows, of course. I'd say it's really been Apple and Adobe, and to a lesser degree Sony, really trying to build a walled garden. And WMDRM has always had RAND licensing; Apple could have licensed it and implemented it themselves if they were so inclined.
DRM was more about liberating the bits from the discs and the fixed-function CE devices. The real competition for WMDRM was DVD and CSS; content provdiers had to feel that their content was going ot be reasoanble secure in order to let it on anything Turing Complete device. The studios giving up on DRM isn't a loss for anyone, I don't think, but a win for anyone making file-based playback devices.
I don't know if it's possible to make significant, sustained revenue from DRM. It's one of those ecosystem components that unblocks a lot of other stuff, but doesn't appear to have have financial value it itself.
I doubt they paid that much. Like On2 with VP6, Sorenson's model was to license the decoder and a consumer grade encoder on the cheap, and then make the real money by selling professional grade encoders once there was a good installed base of players.
Since I said CPU, I thought I was clear that I meant decoding in software.
Sure. But no device meant for media playback is going to use software decoding. It'll already have a decoder ASIC, typically with at least H.264 Baseline and VC-1 Main Profile. Either will offer better quality and much longer battery life in an ASIC implementation compared to a software Theora decoder.
Also, about the battery life -- I can get about 5-6 hours of playback time decoding purely on the CPU with an Atom N280.
IS that with a GMA500? That's capable of hardware accelerated MPEG-2, VC-1, and H.264 decoding.
What are your mobile devices, and what's your media player? All the current ones that are meant for media playback include H.264 ASICs. And those are getting crazy good; the Zune HD's going to support 720p HD playback using the NVidia Tegra.
Lacking an ASIC, any Theora on devices would need to be done in sofware, and even a simple codec can be extremely taxing on a 400-600 MHz ARM. Even if it's playable it's going to eat battery like no tomorrow. I can imagine a really good implementation being able to maybe do 320x240 30p 500 Kbps Theora in software on a 600 MHz ARM, but that'd require a whole lot of tuning.
The VP3 bitstream predates device media, so I doubt it has any particular design tuning for them. It's very much a codec designed for x86, with a PPC port.
Well, YouTube has three sets of settings:
Low bitrate H.263 + MP3
HQ bitrate H.264 + AAC-LC
HD bitrate H.264 + AAC-LC
The low bitrate, for whatever reason is keeping to the specs they've been using since launch, which are using the xvid implementation of old Sorenson Spark H.263 v1/MPEG-4 Part 2 Short Header. Maybe for device compatibility? Anyway, That's a codec about as old as the Theora bitstream, so we wouldn't expect it to be much better.
But I don't know that YouTube thinks it's "good enough" - they're offering higher quality modes, and that's what you get by default on the iPhone and other platforms. For whatever reason they're keeping around a legacy version, likely backwards comaptibility with some clients that don't do H.264 for whatever reason.
For the their high quality streams, Theora isn't competitive in quality. And for the highly compatible streams, Theora isn't competitive in compatibility.
So YouTube saying that Theora doesn't make sense for them makes sense to me. Therora doesn't an advantage in quality or compatibility for the streams they're doing.
Also, Big Buck Bunny isn't the best clip to extrapolate from, as it's really high quality lossless animation. To really see what YouTube needs to handle, try some lousy webcam, DV, and VOB rips. That's where H.264's in-loop deblocking filter give it a big advantage over other codecs, because it just gets smoother intead of blocky as the content gets more challenging.
Not to dismiss the excellent development work Xiph has done on Theora. The posts have been a fascinating read. But it's not plausible to me that anyone can make a business case for Theora over H.264, VC-1, or ASP licensing is available; the reduced bandwidth costs would be bigger than the actual real-world licensing fees for the real world examples I've thought of.
Theora's sweet spot would be in cases where MPEG-LA codec licenses simply aren't available for whatever reason. I imagine a fully refined Theora decoder would need fewer MIPS/pixel than H.264 High Profile, and perhaps even Baseline. But even in those cases, VC-1 Main Profile will probably offer similar performance with significantly better efficiency.
Does Ogg even have hardware acceleration at this point?
Nope. I don't know if anyone's even scoped a hardware implementation of VP3. There have been some VP6/7 DSP implementations, but no ASIC ones (ASIC have better power consumption).
Now, Theora is a pretty simple codec, so doing it in hardware would be a lot simpler than H.264 and probably simpler than VC-1. But it can take quite a few engineering years to refine a decoder for performant playback.
Of couree, performance isn't just the video decoder. It's the video and the audio decoder, and the whole pipeline to make sure you get smooth in-sync playback. Media pipelines are really hard, particularly if you're trying to implement them as part of a browser rendering model that never had to worry about timestamps, decoder buffers, etcetera.
Well they're not pushing it too hard anymore, but that's really because they already lost on the audio side. Their hopes for locking up online music sales died when the major labels agreed to sell without DRM. Video may not be all that far behind.
Once the labels don't require DRM to release content, than the PC and devices are a healthy market for commercial music. Mission accomplished.
Having a good DRM still makes other business models possible, like ZunePass, and it's not like there's some other braodly implemented standards-based DRM as an alternative.
Generally the stuff Windows Media gets used for exclusively is stuff where there's not a viable alternative.
It hurts Microsoft, but particularly because their formats are more open than the proprietary alternative Microsoft is pushing.
Microsoft really isn't "pushing" Windows Media that much anymore. Zune and Xbox already support MPEG-4 and H.264, as will Silverlight 3 and Windows 7.
Really, the money is in having a good flexible media platform much more than in owning one. Windows Media exists to solve problems that weren't otherwise being solved. But the chair of the committee that designed H.264 has been a Microsoft employee for the last decade.
The only company that really made a lot of money on media platform licensing has been On2 with VP6, and their high licensing costs (like multiple $K per core for enterprise encoding) were one big reason that Adobe adopted H.264.
Well, it's not like YouTube sells more ads with better looking video, and I doubt 90% of the uploads get watched more than a dozen times. They probably have some pretty deep metrics about the watts/cents per minute of video encoding and tune for that.
YouTube is also really only a good example of YouTube, since they're a massively money-losing operation funded by a very rich company. No one else does it like YouTube, and ever other video site is going to average a lot higher views/clip, so they can afford more CPU time to improve quality.
Or maybe they're just not very good at video compression :).
Beyond B-frames, they're not using 8x8 blocks or CABAC entropy coding, both of which can offer substantial efficiency improvements.
In recent years, they have both been proponents of DRM at some point, both support their own proprietary formats (Microsoft with WMA/WMV/ASF
The WMV 9 video codec has been standardized as VC-1 via SMPTE. Its licensing is handled by MPEG-LA, like MPEG-2, MPEG-4 part 2, and H.264.
Apple with QuickTime, and AAC
QuickTime was the basis for and largely the same as the MPEG-4 file format.
AAC==Advanced Audio Codec. It's a MPEG technology originally created for MPEG-2. Apple is one of many implementors, and was not the primary developer (which was AT&T, I believe).
Microsoft won't support a codec that makes the web more reliable for non-Windows users - especially Linux users.
Actually, the currently in beta Silverlight 3 supports WMV and H.264, and has an extensible Raw AV pipeline that would make implementing the Ogg codecs inside of Silverlight trivial.
And that technology is already impleented in Moonlight, so Linux users have access to it as well.
So, Microsoft is actively exactly doing that, and includig Mac and Linux users.
Theora is a pretty weird codec. It's got a strange scan pattern, goes from bottom up instead of top down, and has lots of other rather unique aspects that could make a silicon implementation harder than its relative simplicity would suggest.
The funny thing is there are royalty free codecs out there.
I don't think anyone's ever asserted patents against MPEG-1, H.263, or MP2 audio, and any patents would be due to expeire soon anyway.
If being patent and license free is the paramount concern, there's plenty of choices with much more mature implementations than Theora. I don't understand the exclusive focus on just a few immature codecs like Theora and to a lesser degree Dirac.
Now, the bigger challenge is weighing the cost of patent licensing versus the cost of bandwidth. H.264 pencils out cheaper than Theora in practice when bandwidth costs and audience size are compared (lower bandwisth==bigger audience with sufficient bandwidth for the content).
Theora is based on VP3, which was both patented and commercially distributed for a number of years. If VP3 had been infringing someone else's patents, then they would likely have sued back when a company was making money from it.
Well, I don't believe On2 ever made any money from VP3. VP6 with Flash 8+ was their big moneymaker. They also licensed the TrueMotion codecs for games back in the day, and even licensed their TrueMotion RT codec fof $5M to Microsoft for their never-shipped NetShow Theater product.
But it was pretty much the DotCom investment bubble that floated them from the late 90's through Flash 8. And they shrunk by a good 90% in staff from 2000 until VP6.
I met with them at a trade show a few months ago and realized I'd actually been having business dealings with On2 (to back when they were called the Duck Corporation) for longer than any current employee has been there. It's weird telling a company their own war stories back to them :).
By Windows Media do you mean DirectShow? DirectShow is extensible to 3rd party codces, and there are Theora implementations. But not for Windows Media.
There are really only two significant video formats today for web streaming: Mpeg4/H.264 with MP3 or AAC audio is technically superior; Ogg/Theora with Vorbis audio is freer.
Ogg use on the internet is a rounding error at best; RealMedia still gets more use (very popular in internet cafes in China for some reason).
The three primary media formats/codecs are MPEG-4 + H.264 (QuickTime, plus Flash and soon Silverlight 3 via progressive download), RTMPe + H.264 (Flash uses MPEG-4 files but a propritary protocol), and Windows Media + VC-1. Move Networks + VP7 (ala ABC.com) also pulls in million of eyeball/hours a month, certainly more than Ogg at this point.
I'd say Ogg is #5 at best today. #6 if you count torrents and hence MPEG-4 part 2.
As for Microsoft support, that's becoming pretty codec neutral. Silverlight 3 (currently in beta) supports both H.264 and has a Raw AV pipeline allowing arbitrary codecs in managed code to be added to any Silverlight player. So adding Theora/Vorbis or any other codec, format, and protocol can be done inside the Silverlight sandbox by any third party.
Ignoring the tremendous improvements in the Thusnelda branch, if YouTube suddenly switched from severe H.26whatever overcompression to stock Theora with optimal settings (and everyone had libtheora and HTML 5 browsers), no one would notice the difference.
Untrue. Xiph has made heroic progress with Theora, but it's still a decade-old codec design and bitstream, and it's hard to imagine it catching up with xvid, let alone a good H.264 implementation.
YouTube certainly has quality issues, but things can be bad in more than one way at a time. There's nothing that less efficient codec would help them with. Note their top bitrate is 1280x720p30 at 2 Mbps.
Some samples compared Xiph's latest demo clips, with the same source encoded with VC-1 and x264 are here:
http://cid-bee3c9ac9541c85b.skydrive.live.com/browse.aspx/.Public/BBB%7C_Compare
x264 can do 640x352 with higher per pixel-quality than Theora can do at 400x224 at the same bitrate.