Working With Ogg Theora and the Video Tag

An anonymous reader writes "The Free Software Foundation's Holmes Wilson is just back from Berlin, where he participated in the Ogg Theora book sprint put on by FLOSS Manuals. Here is a broad look at Ogg Theora and how it fits into the push for free formats: where we're winning, what works, and what could be improved."

Comments

Theora

[ardor]

Unfortunately, Theora will stay irrelevant where it matters most. In sites like Youtube, h264 will prevail. And this time, h264 is the (much) better tech as well.
To get the same quality as h264 video, Theora video needs higher bit rates, which translates to higher traffic, and in the end costs more money. The much higher popularity of h264 compared to Theora doesn't help, either.

Re:Theora

[Anonymous Coward]

Agreed. And the OP linked article has a joke-of-a-comparison. I encoded the same video, same dimensions, same frame rate, and was able to widdle h264+AAC bandwidth down to 260 kbps and it still looked better than Ogg/Theora+Vorbis especially where the scene zooms towards the dark cave with sleeping bunny.

Re:Theora

[hairyfeet]

Not to mention the lack of hardware acceleration makes it pretty much a non starter. My graphics card that cost a whole $50 (a 4650) came with H264, WMV9, DivX, and MPEG 2 & 4 out of the box. And with the rise of netbooks/nettops, green computing, mobile devices and high def video now more than ever hardware acceleration is the way to go. Is there even a beta driver for Theora that gives ANY acceleration?

Without hardware acceleration, preferably given to the big three (AMD, Intel, Nvidia) so they can integrate it into their drivers so users can get full acceleration easily and out of the box, I just don't see Theora gaining any ground. I know that those that support FOSS find this hard to accept, but Joe user really doesn't care if a codec is free or not, hell most don't even know what a codec is, they just want easy to use and simple. Theora need hardware support like yesterday if they want to gain traction. Although ultimately I think it will be like Vorbis VS MP3. Vorbis might work fine, but my MP3 player doesn't play Vorbis, in fact the majority don't. Folks don't care that MP3 is encumbered because it works for them. So while I wish the Theora guys luck it looks like a seriously uphill battle from where I'm sitting.

Re:Theora

[Anonymous Coward]

And staying with that kind of thought process, one wonders why anybody bothered with Linux development from the mide 90's.

Re:Theora

[FrostedWheat]

Not to mention the lack of hardware acceleration makes it pretty much a non starter

You say this, but nowhere do you say why it needs hardware acceleration. Have you even tried it? My fairly old machine plays a 1080p Theora video just fine. A completely unscientific test with top shows about 33% CPU usage, peeking at about 40%. The same machine cannot decode 1080p H.264 video in real time.

Theora just isn't as CPU greedy as H.264 -- it doesn't need hardware acceleration. Although it wouldn't hurt ;-)

Re:Theora

[skribble]

Hardware acceleration (and dedicated hardware in general) is much more efficient (and for manufacturers, much less expensive) then general purpose processing. So while your computer may have a processor that can handle this, many smaller consumer devices don't... additionally, for portable content you need energy efficiency... how long would you computer processor run on a cell phone battery? Also... hardware acceleration isn't just for play back. It's also for video creation/production. Many pro video systems take raw video and encode it to h.264 on the fly in real time (For SD/HD streaming and well as broadcast distribution). And then there's other studio productions where hardware acceleration allows working with real time effects... etc. The reality is that playback is a very small part of the puzzle... If you want to push Ogg-Theroa as a standard then you need the product creators to use this, and... there is no compelling reason to do so. I support much Open-Source software (both with my time & contributions and direct financial support)... that said, in the real world you pick the best solution for the problem, and in this case Ogg-Theroa is not it. And this is no disrespect to the development of this... this is hard stuff to do and it's really incredible that Ogg-Theora is where it is today, unfortunately it falls short of h.264. Also... pushing an inferior standard down the throats of a web viewing public, isn't going to win the open source model any friends.

Re:Theora

[TheTurtlesMoves]

The relevance of Theora was and is like the relevance of ogg. Sure its not going to take over. But without it, the other codec owners (apple, MS etc) would have no "license free" completion to keep them honest. Instead of a debate about licensed and unlicensed codecs, it would have been a debate on how the hell could we afford these kind of license fees to claim any kind of standard in the first place.

As for the higher bitrates arguments. Honestly you tube and co look like such crap I can't believe people ever bother, theora dose not look worse that these bit rates. I have decided most people are in fact blind.

Re:Theora

[mugginz]

One thing for sure, hardware accel for Theora will only come after (if) there is wide adoption of the format. The only likely caveat to this that I can see would be Theora decode via OpenCL produced by the community.

Re:Theora

[mugginz]

With any luck, the findings pointed to by http://web.mit.edu/xiphmont/Public/theora/demo8.html may lead to better quality/bit-rate ratios in the future.

Re:Theora

[CSMatt]

I am sick of this argument. Not because of its merit, but because of its defeatist attitude. Arguments like this one are basically "Theora has no hardware decoders at the moment, therefore it never will, and Theora will die." They may not be saying this explicitly, but the implication is certainly there.

At what point do you think that someone won't step up to the plate and design a Theora decoder? I don't know of any technical reason that they can't. The decoder has been frozen since 2004, so encoder improvements can still continue and will still play on hardware-accelerated devices. The only thing that is needed is sufficient demand for the codec. The codecs you mentioned as now having hardware accelerators started out without any. It was only after those formats became popular that they started being built.

Furthermore, mobile devices may or may not need these decoders, but the overpowered Core Octo machines that the soccer moms and grandmothers of the world are being told that they have to buy for their e-mail and word processing needs will handle Theora without any acceleration. You mentioned your $50 card? Well, my card has no acceleration at all, and I can play web-resolution Theora just fine. In fact, I can play multiple videos at the same time.

Defeatist arguments like this one aren't helping. They are hurting. They only serve to create self-fulfilling prophecies by discouraging Theora adoption on the grounds of something that will never come to pass unless Theora adoption occurs in the first place. If you really do want Theora to succeed, then it needs to grow enough of a base that chip manufacturers will see a high enough demand for Theora chip production.

Where are we winning?

[BadAnalogyGuy]

FTFA: Ogg Theora is becoming a big deal

I have worked at various companies, from small ventures up to well-known large corporations and have found the same thing at each. Employees think that their company is pretty well-known in their respective fields. While it may be true of some companies (IBM, Microsoft, Amazon, Johnson & Johnson, just to name a few), most third party vendors are mere gnats on the backs of those wildebeests.

This is myopia caused by too much focus on a specialized area. Yes, maybe within a very limited sector your technology may be making inroads, in general you are nothing more than a butterfly flapping its wings. Theora is not becoming a big deal. It is just another codec, and one that isn't particularly popular.

There are technical issues that need to be addressed technically, not simply (as the author of the article does) waved away as a future feature to be implemented when the codec becomes more popular. It will never become more popular until it can offer sufficient reason to switch. Relying on the negative influence of patent encumbrance to drive people towards the codec is a losing proposition. It is a reactive strategy that cannot eventually win.

What struck me most about this article was how even the FSF is not particularly behind Theora, per se. They are for "patent unencumbered" codecs, so they have no real inclination to push Theora in the marketplace. Without a proactive strategy to push Theora both in a business sense as well as technically, it will flounder.

Another codec bites the dust. Big deal.

Theora 1.1

[Torrance]

Monty from Xiph has provided an update on the state of the upcoming 1.1 release. It makes for interesting reading.

The bigger picture

[malevolentjelly]

I don't think there's any evidence that the video tag is catching on in any meaningful way. Can anyone point me to evidence of the contrary?

Who is to say that Flash's grasp is even weakening among major content providers? Is the video tag DRM friendly?

Re:The bigger picture

[MartinSchou]

Well, that was anti climactic.

Opera 10 beta 3: Shows the player, but doesn't work "You must have an HTML5 capable browser."
Firefox 3.5.2: Shows the player, but doesn't work. Doesn't give the error message though
Google Chrome 2.0.172: Same as Opera "You must have an HTML5 capable browser."
Google Chrome 3.0.195.6 (latest beta): All player controls work except full screen and the thingie on the right hand side, but none of the "more from" or "related videos" links work at all.
Internet Explorer 8: Only shows the controls for the player, "Done, but with errors on page"
Apple Safari 4.0.3: Can play the video (yay), but nothing else works. Doesn't show the time played or remaining, doesn't move the time indicator, none of the "more from" or "related videos" links work at all.

I've no idea if the issue is with YouTube or with the browsers, but ... it's really not impressive. I installed the latest Chrome beta just to see if that made everything work like it should on that page, and it still doesn't.

I've no doubt that it will work eventually, but for now, I wouldn't use that site as an of course it works, just look at this example.

Technically inferior

[Anonymous Coward]

The main problem with Theora, is that it is clearly technically inferior.
For instance, Vorbis generates comparable or better quality than MP3 of the same size, so it has a hope to be pushed. Theora doesn't.

Re:hmmm

[beelsebob]

The reasoning is that they wanted to put one codec in the spec that they could guarantee that all vendors would support, roughly like flash is now through plugins. That way, the video tag would actually be usable, website authors could guarantee that unless people used crazy browsers from crazyville, they would be able to watch the video.

In the mean time though:
â Mozilla refused to support h264 because it is patent encumbered.
â Apple refused to support ogg because it's technically inferior and they didn't want to put dev effort into something worse than they already have.

The result was that no decission could be made on which one would be supported everywhere.

Re:Google behind HTML5... Not behind Theora

[David Gerard]

Yes, and when he was called out on his BS and FUD ... he promptly disappeared.

GUI editor for Theora: LiVES

[AliasMarlowe]

Try finding video editing software which can edit (not commandline like ffmpeg, I'm talking gui After Effects style) a Theora file.

I've never used After Effects so I'm not sure what features it has. However, if you want a GUI editor which can handle theora files, then try LiVES. It's rather better (in features & interface) than avidemux or kdenlive, neither of which can handle theora. It's cross-platform OSS for BSD-Linux-Mac-Windows.
http://en.wikipedia.org/wiki/LiVES http://lives.sourceforge.net/

Re:Subtitles

[FrostedWheat]

That's OggKate's job. It also works with any other Ogg embedded video codec.

Compare success of Web vs audio video standards

[gig]

Audio standardization is not only bigger than the Web, it's older, and it's MUCH more successful than any Web standardization to date, including HTML 5, which is still only 35% of desktops and 90% of mobiles.

I think until the Web development community actually creates and follows even just one of their own standards (maybe HTML 5 will be the one), browser makers and other principals should STFU about audio video standardization, which has been highly successful for 30 years.

During the 21st century thus far, you can't make one fucking Web page for all browsers. But the same ISO MPEG-4 audio video plays in both Adobe Flash and QuickTime Player; both iTunes and YouTube; both iPod and Blu-Ray; both iPhone and Blackberry. Camcorders from Sony and Kodak make the same MPEG-4 video format. Editors from Adobe and Apple edit and export the same MPEG-4 video format. Both NVIDIA and AMD GPU's have ISO MPEG-4 H.264/AAC decoders in them. There are MPEG-4 players from literally hundreds of manufacturers.

But consider that Linux and Windows can't play all of that audio video, and so we invoke Flash in a Web page, bring in a proprietary app with questionable security context and crashy history and also it changed owners twice already, just so we can make everyday standard audio video work in Linux and Windows!

And during the 21st century thus far, HTML has been static. The object tag bullshit from 2008 is the same object tag bullshit I used in 1998. The W3C and browser makers have contributed almost nothing to audio and video in the entire history of the Web. If not for the fact Tim Berners-Lee created the Web on a NeXT system that had 8-bit audio, maybe the Web would not have had audio at all from the beginning. The Web is turning 20 and still no consumer level audio, never mind pro audio. I produce music ... how can I express a 5 minute 24-track 24-bit 192kHz song made up of hundreds of synchronized audio clips in HTML so I can store it for posterity? You guys are not even getting started with what needs to be done with audio and video on the Web. And while HTML did nothing over the past 10 years, we got RSS and then podcasts, which are filled with ISO MPEG-4 audio video. Even MSNBC.com is MPEG-4 since podcasts, no more Windows Media. Set-top boxes with MPEG-4 decoders in them are downloading podcasts. These podcasts are viewable already in browsers. The browser today is interacting with a metric shitload of MPEG-4, but it's leaving it all to Flash and then ironically, the browser vendors complain that Flash crashes their browser! Incredible.

Think about the fact that Microsoft could not break MPEG-4 standardization in spite of using Windows and Internet Explorer to push Windows Media. That was years ago when MPEG-2 was changing over to MPEG-4. How is Firefox going to do it now, when all the media is MPEG-4 already?

Understand that music and movie makers are creating content for MPEG-4 in the way they used to make CD and DVD. Authoring tools have had MPEG-4 export for many years, it's extremely old news. And music and movies are not tolerant of format wars. The margins are too low. Most music albums and movies don't make money. A format war kills all the smaller artists who can't double up their production costs to make 2 products. Broken audio video standardization breaks artists. The media that is on iPod and YouTube and Blu-Ray is what is going to be on the Web servers. If Mozilla can't play it then Flash will be invoked perpetually. That is all Flash is used for now it seems, is to wrap MPEG-4 up to make it Linux and Windows safe.

Further, this is all political because there is no technical substitute for MPEG-4 that pleases Mozilla. Ogg is offered, but Google has already said that an Ogg YouTube would require more bandwidth than currently exists in the world today. Are you telling me that YouTube is not part of the World Wide Web? Ogg on iPod would get you one quarter battery life because there are no Ogg decoder chips. Should the audio from the Web not play on iPods

Re:... wait we already lost!? When did that happen

[TheRaven64]

Well, the MPEG-LA is doing a good job with their plan to introduce per-download fees for people using H.264 next year. If you're still using H.264 for streaming video next year, for anything longer than a 10-minute clip, expect to be giving all of your profits away to the MPEG-LA. Or you could switch to some other CODEC like, for example, Theora, which doesn't have stupidly-expensive licensing fees.

To be honest, I'm more interested in Dirac than Theora. VC-2 is a profile of Dirac which, like Theora, is not patent-encumbered. It's based on wavelets and is much higher quality and has a lot more industry backing than Theora (the BBC, for example, are using it for archiving already). Currently, the CPU requirements for decoding Dirac are a bit high. Playing back the Big Buck Bunny example on my 2.16GHz Core 2 Duo uses 100% of one core (although I'm using a slightly old version of the CODEC, apparently the latest one is about 20-30% faster). The BBC is working with hardware manufacturers to get hardware decoders which should make it a lot more attractive. There's also a CUDA-based implementation and a GLSL version which are reported to be a lot faster than the CPU-based version (I've not tried either) and should work on most modern GPUs. Given that most modern handhelds now include an OpenGL ES 2.0 GPU, which means that they support GLSL, it's likely that Dirac playback on handhelds will work nicely soon.

Theora has much lower CPU requirements than even H.264, so using Theora for the low-quality version and Dirac for high quality sounds like a sensible approach.

Re:hmmm

[TheRaven64]

Submarine patents are not a valid reason for choosing H.264 and not Theora. The VP3 codec that Theora was based upon has been around for longer than H.264 so there has been longer for patent trolls to come out of the woodwork. The H.264 license you get from the MPEG-LA doesn't grant you a license to all of the patents required for H.264, it grants you a license to all of the patents that the MPEG-LA knows about required for H.264. Similarly, the (free and irrevocable) patent grant from On2 gives you a license to all of the patents that On2 knows about required for Theora.

Oh, and it wouldn't take a few hours for an Apple developer to add Theora support, it would take zero hours. On my machine, video tags referring to Theora content Just Work(tm) in Safari. Safari doesn't do any video decoding, it delegates it to QuickTime. If you have the (free, provided by Xiph) QuickTime Theora plugin installed then Theora videos work correctly. Installing the codec isn't exactly difficult and once you've done it you never have to think about it again. The problem is on the iPhone, where users can not install their own codecs.