OSNews

I only listen to real lawyers on patent matters.

http://tieguy.org/blog/2010/03/25/patent-101/

Everyone else is just masturbating

Update:

http://tieguy.org/blog/2010/03/26/more-patent-101-and-some-patent-l...

It has Google on its side.

Yes, but Google use H.264 as well. They include OGG in Chrome for the simple reason that, as an open source product, the H.264 component will not be available in all instances and all locales (particularly Linux). By providing both, Google allow Chrome a fallback mechanism under the assumption that developers will already be providing an OGG encode for Firefox support.

If legitimate, quantifiable patents rise against OGG, then what's to say that Google won't just simply drop OGG from Chrome.

Only their purchase of On2 demonstrates their concern against H.264. OGG has no patents to protect itself, but if Google own On2, then they own a bunch of patents which can be used a defensive collateral for VP8.

I'm playing devil's advocate anyway. I agree with you, and thank you for the well written article.

MPEG-LA licenses seem to be limited to the specific codecs in most cases. I don't know the license Google acquired, but with your average "I can use H.264" license, you didn't license the patents for any other use.

Agreed. (couldn't +1 you)

Also, the fact that Youtube favours H.264 over Theora should be warning signs enough for people who want to speculate over Google future codec allegiances

It's a bit early, but that's why h.264 has probably already lost. I laughed at all those people spouting nonsense regarding Firefox and "Oh, if they don't support h.264 no one will use them". Well, it's the other way on - Google knows Firefox is out there, not everyone can play back h.264 and they simply have to provide a fallback.

In then end, people are just going to use the more widely available format and, paradoxically, browsers and devices are going to use the format that is more widely available and that they don't have to pay ridiculous fees for - which could change at any time, regardless of assurances.


Ifs, buts maybes. We know h.264 is patented and you have to pay for the thing. We know of no way that Theora is currently patented. That's the situation, and the latter is infinitely preferable.

Google won't drop Theora support from Chrome because history has shown us that once you've started supporting something you can't just drop it that easily. They've provided Theora support for the reasons you've stated above, and to go back on that is rather messy.


Quite frankly, I'm not interested in Google protecting us. Google already support Theora, and as such they will have to move to protect it in exactly the same way because they will have content in it and users who need to play it back.


Always nice. Having gone through it I'm more certain that h.264 can't have a future as a standard on the web, and fortunately we have an alternative which we didn't when we got hoodwinked with GIF.

Thing is, most browsers can playback h.264, since they have a flash plugin. It's possible to determine in JavaScript whether a browser's video tag will support h.264 or not, so one would simply replace video tags with a fallback flash player.

Thus Google can keep YouTube serving out h.264, and not bother encoding Theora versions for Firefox.

What I think is most absurd about Firefox's position is that they're not just handing off all codec handing to an external framework. Safari happily supports Theora and Ogg streams if you have the Xiph QuickTime components installed...

Flash is not h.264.


I'm pretty sure you can't actually do that with Firefox - and Firefox has done that deliberately.


Nevertheless, Google is supporting Theora.


No one cares about third-party codec support. This is all about a lowest common denominator as to what every browser can feasibly support - both technologically and legally.

We went through all that third-party codecs stuff for many years with Real and Windows Media and video on the internet. It failed because no one could reasonably rely on what would be installed and supported. That's why Flash video has become rather a defacto standard. "Oh, you can install a third-party codec or hand off to a framework...." does not provide reliable internet video and is not an answer.

I'm pretty sure you can't actually do that with Firefox - and Firefox has done that deliberately. "

Maybe not directly, but I'm willing to bet you could just detect the driver version, and do A if Firefox and B if anyone else. Or do h.264 if (browser_id is in h264_browsers), and flash_fallback otherwise.



Well-said and good point. What I kinda don't get is why we can't just have multiple blessed codecs (much as we have multiple blessed image formats)?

And why may I ask would Google remotely give a shit about Theora? in all seriousness - why would they give a crap considering that it would be simply a matter of dropping Theora support from Chrome? You're trying to make out that Google actually cares about Theora when in reality its nothing more than a lip service to the open source crowd in the same way that Darwin (Mac OS X) is nothing more than lip service to the open source world.

Edited 2010-03-25 12:52 UTC

And why may I ask would Google remotely give a shit about Theora? in all seriousness - why would they give a crap considering that it would be simply a matter of dropping Theora support from Chrome? You're trying to make out that Google actually cares about Theora when in reality its nothing more than a lip service to the open source crowd in the same way that Darwin (Mac OS X) is nothing more than lip service to the open source world. "

It is about control. Control of "permission" for Google to serve video on the web.

If Google use Theora, then Google have control over Google's ability to serve video on the web. No-one can hinder Google's right, or ability, to do it, if they use Theora.

If Google use H.264, then these companies would have control over Google's permission to serve video on the web:

http://www.mpegla.com/main/programs/AVC/Pages/Licensors.aspx

That isn't exactly a list of "friends of Google".

I think you may find that patent holders do not have to be "reasonable". They can simply say to Google "no, you may not serve h.264 video, because we don't like you".

If all of the web clients can only render h.264 vide, Google would be out of the game. Fortunately, we are nowhere near that point yet.

Edited 2010-03-25 13:02 UTC

Because a patent suit against Google regarding Theora wouldn't be solved by that at all. Theora has been included in Chrome for a while now, and any patent troll would want to see damages incurred during that time.

Dropping Theora support would solve only half of the equation in this hypothetical scneario.

Because a patent suit against Google regarding Theora wouldn't be solved by that at all. Theora has been included in Chrome for a while now, and any patent troll would want to see damages incurred during that time.

Dropping Theora support would solve only half of the equation in this hypothetical scneario. "

That is assuming a patent troll would win. Theora is based on a pretty old codec. I'm pretty certain (by now) that there is no relevant, earlier, patented technology out there. Apple have been looking for it for nearly ten years now, and haven't come up with any.

I'm sorry, I still need to get over the fact we're on the same side on this one.

I keep expecting Elvis to land in his space ship to balance it all out.

Because a patent suit against Google regarding Theora wouldn't be solved by that at all. Theora has been included in Chrome for a while now, and any patent troll would want to see damages incurred during that time.

Dropping Theora support would solve only half of the equation in this hypothetical scneario. "

That is assuming a patent troll would win. Theora is based on a pretty old codec. I'm pretty certain (by now) that there is no relevant, earlier, patented technology out there. Apple have been looking for it for nearly ten years now, and haven't come up with any. [/q]

Apple's not "looking for patents" they paid up "protection" to MPEG-LA and at this point they are the 800lb gorilla in the room. I'd venture Apple is getting the royalties for much less than anybody else... because they control enough media market, where Apple goes, the market goes. h.264 helps with lock-in to their stack... Apple wants to help "creatives" but only with "paid for", "locked down" tools.... Most OSS codex are locked out of Apple by default... they don't even toss in an EXT2/3 driver or other pieces that would play nicer with Linux and gain allies. Apple's biggest weakness right now is trying to "walk their own road" and not being willing to bring in other folks "enemy of my enemy" type deals.

Hypothetically, I think it (dropping Theora support) could at least reduce the damages that might be awarded. IANAL, but I think that, using loose language, the more of a good-faith effort you put forward to resolve a situation, the lower your damages are likely to be.

I also agree with Kaiwaii: I don't think Google are deeply committed to Theora in Chrome. It's more likely that they realize that, given that Chrome is technically an open-source project, they'd have to do extra work to keep us freetards from shoving Theora support in there somewhere. I strongly suspect they're just appeasing the FOSS community, more than they really care about supporting Theora. I think the comparison to Darwin is actually fairly apt.

Google is the boyfriend that smacks the FOSS crowd around a little when he gets drunk but is still a "good guy" since there was that one time that he stood up for her at the county fair.

As for patent trolls Google isn't a good target since they aren't making money off the Theora codec. If they were sued they could just claim ignorance and remove it.

But let's say for the sake of argument that Theora does violate patents. Whose patents would those most likely be? MPEG-LA, and Google doesn't have to worry about their codec alliance partner suing them for including an open source codec that only exists in their browser for image reasons.

Are we forgetting Summer of Code? It's basically Google paying students for participating in FOSS projects. This year it'll be the 6th time they are doing SoC, so "since there was that one time" is just misleading or an outright lie.

It isn't misleading or a lie because it's a half-joking in the first place. Get sense of humor already.

But yes Google has projects like the summer of cheap labor for appeasing the FOSS crowd and they eat it up. Some of those FOSS advocates even believe Google is an open source company when in reality their business is based around a highly secret proprietary search algorithm. The fact that they don't even release their own internal build of Linux says enough.

I can't believe how strong the FOSS / Google love affair is after they stonewalled Theora. But I suppose a company with billions that doesn't like Microsoft will always turn on FOSS advocates. They probably can't help themselves much like a woman who is sexually attracted to a jerk.

Meh. Anyone is allowed to write any software for their own purposes, and do with it what they please. It is their software.



Read the GPL license.
http://www.gnu.org/licenses/gpl.html

The copyleft provisions of that license apply ONLY to re-distribution. It is one of the four freedoms that you are guaranteed with FOSS software that you may modify the code as much as you please and use it for your own purposes without having to reveal your changes to anyone.

http://www.gnu.org/licenses/quick-guide-gplv3.html


The first two freedoms apply here. These two say NOTHING about a requirement to share you changes if you use the software for your own purposes.



Why do advocates of proprietary software constantly feel the need to unfairly disparage others? What is WRONG with these people?

Edited 2010-03-26 02:20 UTC

Yes that is a bit odd but and so is all the love for IBM.
Btw, that boyfriend thing? That was funny.

A half-joke is 50% truth. It can just as easily mislead or lie as any other statement, although it tends to be harder to spot. That's why people love metaphors to make arguments, since they rarely get called out for cherry picking what parts of the metaphor actually apply.

OTOH, a half-joke is 50% funny as well. I'm not quite sure that a quick reference to Stockholm syndrome is really amusing, since it's not an intrinsically funny topic. I suppose it's cynical, but cynicism is a distinct entity from jest.

If google really believes that supporting Theora is minimal threat, why don't they Use it on YOUTUBE? why did they invested in buying a new Codecs? Google is like a whore if you pay them (with supporting Theora, they get the OSS- Supporters heart), they gonna do what ever you want! They do any thing to get more users and Ads market.
Only the future will tell, I don't believe that Mozilla will keep their Position, Google is taking their Market-space. Buy supporting both Codecs they are getting Firefox-users, but they will not use Theora in any Productive Space.

my two cents

As long as MPEG LA do not control the market completely, MPEG LA need YouTube to be h.264. MPEG LA might even SPONSOR YouTube to use h.264 ... that is a possibility.

Eraly last year, MPEG LA were starting to get very confident of their position. There was no other competitive codec. Mozilla announced a donation to improve Theora, but pfffft ... that wasn't a threat (or so thought MPEG LA).

http://techcrunch.com/2009/01/26/mozilla-gives-100000-grant-towards...

Early last year, MPEG LA began talking about increases in the license fees for h.264 for use on the web. Google saw the writing on the wall, and bought On2 (about mid-year), hoping to gain an alternative codec from that.

Unfortunately, On2 are licensees of MPEG LA.

http://www.mpegla.com/main/programs/AVC/Pages/Licensees.aspx
(number 465).

Late last year (well after Google had bought On2), Mozilla's funding of Xiph.org for development of Theora bore fruit. Xiph.org released the Thusnelda branch of Theora, which finally was competitive with h.264.

As a side-note, development of Theora continues, and the next branch (called ptalarbvorm, which is now still experimental) is an appreciable improvement over Thusnelda (and hence, an improvement over h.264).

http://people.xiph.org/~greg/video/ptalarbvorm/

Thusnelda and Ptalarbvorm both make perfect sense for Google to use "in production space". The earlier Theora 1.0, which was out when Google bought On2, was not suitable.

Edited 2010-03-25 13:38 UTC

It has Google on its side. "

And Google is the antithesis of evil... right?

It has Google on its side. "

And Google is the antithesis of evil... right? [/q]

What does that have to do with anything? Gruber claims nobody stands behind Theora - but cleary, Google does. That's all there is to is. Your point about Google being evil is irrelevant.

And you're obviously right. Like any other company, Google is inherently evil. Just in this particular matter, they're on the good side.

And Google will defend Theora, in case of litigation, instead of dropping it immediately from Chrome like a hot pile of shit, because?

Google is already licensed to use H264, has already implemented H264 in Chrome in addition to theory, and is a heavy user of H264 in, you know, THE WORLD'S BIGGEST VIDEO SITE.

Chrome's use of Theora is mostly for completeness and to get some good vibrations from the FLOSS community.

I suspect Google is just supporting Theora because it's what Wikipedia is insisting on and not everybody has installed Java for their fallback player.

It has Google on its side. "

more importantly the Theora codex does have patents filed from the original authors and signed over to xliph. The patented codex is known to all the competitors. Unfortunately, there is little in the law about disclosing when your patent is violated... even if you're directly asked. It's "public knowledge" because it's in the Patent database....but if you search the database then you become liable for triple damages, yeah!

I do agree there are underhanded ways of getting patents through. Like the MP3 fiasco where companies that WERE part of the original MPEG-LA settlement used technicalities and loopholes to REMOVE MPEG-LA's right to sublicense then started a whole new round of suits against the whole list that paid up once. This is the problem Theora has, and because non-profits don't have deep pockets the money to gain is by waiting for people with money to infringe rather than making their intentions known.

Well the HTC has Google on it's side, but earlier this week, Jobs & Schmidt were seen together having coffee. Apple should just go on ahead & smash HTC to smithereens.

This ignores the fact that the Theora technology is older than h.264.

If Theora and H.264 do have some common technology, and there is a patent squabble, then Theora prevails, not h.264.

No-one else (other than MPEG LA members) are making any grumble about Theora. This has been the case for almost ten years now.

Finally, all that the MPEG LA members ever do is make ominous-sounding-but-vague hints of potential patent difficulties for Theora. All hat and no cattle.

http://en.wiktionary.org/wiki/all_hat_and_no_cattle

Edited 2010-03-25 13:56 UTC

Theoretically, there could be something that VP3 didn't do, but Theora did. In that case, it COULD infringe on H.264.

Also, "H.264" isn't A patent, it's a patent pool. I haven't looked, but there could be patents in the pool that predate VP3.

Theoretically, I suppose. However, I have a player that is Theora 1.0, and it can still play these videos encoded by later versions of the encoder:

http://people.xiph.org/~greg/video/ptalarbvorm/

Xiph.org aren't playing about with the format or mucking about with new technologies, they are simply optimising the encoder.

PS: In threads like these, I am forever astounded by how many people seem to be desperate to try to find a problem for Theora. Theora is a collaboratively-developed freedom codec which anyone may use anywhare anytime in any context without restriction, a gift to humanity if you will.

Meanwhile, the competing codec H.264 is heavily patented, licensors of the codec make endless "all hat and no cattle" threats to sue left right & centre, both against competing codecs and their own users, and they have been doing so for almost ten years, but there is never any actual substance.

Why does ANYONE support those trolls?

1) Because Apple does.
2) Because the quality question hasn't been properly settled yet - as in, the current version might be up to par/better, but without the actual tools using that latest version, people don't encounter said videos, strengthening their idea that Theora isn't as good as H264.
3) Because Apple does.

I performed a simple experiment for myself, that anyone may check.

I download a good-quality h.264 movie trailer.


Here are some screenshots of it playing on my system:
http://ourlan.homelinux.net/qdig/?Qwd=./Mov_480p&Qiv=name&Qis=M

I then downloaded a current Ogg converter, I used the Firefox extension called Firefogg:
http://firefogg.org/

I used Firefogg to convert this already-compressed-by-h264 video clip into Ogg video (Theora).



Here are some screenshots of it playing on my system:
http://ourlan.homelinux.net/qdig/?Qwd=./Theora_480p&Qiv=name&Qis=M

(Sorry, but I'm not quick enough to get the exact same frames as a screenshot).

The Theora video is indistinguishable in quality form the h.264 one, yet it is only 72% of the file size and it has been transcoded from an already-compressed source against which we are comparing it.

The Theora video has every factor stacked against it in this test, but it works fine.

That is a QED as far as I am concerned. Theora has now got h.264 covered, quality-wise.

Apart from the actual tool called firefogg I linked above, here is another actual tool that Windows or Mac users might like to use:

http://www.mirovideoconverter.com/

How about that, tools for open video that anyone may use, even Mac and Windows users. Linux users have been able to use tools for Theora for ages, so I guess it is nice that the closed desktop platforms are finally catching up here.

Edited 2010-03-25 14:50 UTC

I used Firefogg to convert this already-compressed-by-h264 video clip into Ogg video (Theora).

Using it on an already-compressed clip is idiotic. Obtain something that isn't compressed either by h.264 or Theora and then compress it with both and only then compare the results.

Idiotic is your comment.
By decoding an already encoded video and then re-encoding it, the quality can only go south. But in fact it doesn't because the quality of theora's video is as good as the original. Lemur's experiment proves that the quality is the same and theora is on par with h264.
Edit: typo

Edited 2010-03-25 15:09 UTC

I'm sorry, I'll try to be polite, but by saying this you show that you clearly don't understand lossy compression.

By allowing h.264 to encode the uncompressed video, but giving Theora only a lossy copy to work with, I have severly handicapped Theora in this comparison.

Theora still came up trumps.

Edited 2010-03-25 22:32 UTC

If you use mplayer, you can go frame-by-frame by pressing the '.' (period) key, which should solve your problem.

Of course they are indistinguishable! You used a bitrate of over 4 Mb/s on 480p video! That's insane. For comparison a 2 hour movie on a 700 MB CD is about 0.8 Mb/s.

Better luck next time.

No: YOU think he's wrong, and here's why (perhaps).
That's not because it's YOUR opinion that it's the truth...


Correction here

1) Because you hate Apple.
2) Eugenia (and a lot of people) already make articles showing that Theora IS not as good as H264. But you seems to just not read them.
3) Because you hate Apple.

And just because you own Apple stuff doesn't give your some credibility not to be an Apple hater. I am running Windows every day (more than any oses in fact) and I hate it.


Because, in every case, Google has the money to pay for the patents ...

Edited 2010-03-25 15:44 UTC

I don't SUPPORT them.

It's just that I lean towards pragmatist. H.264 has quite a lot of momentum behind it.

My prediction is that, if <video> takes hold, Firefox is dead, unless the plugin approach (the very thing that <video> is trying to avoid,) preferably using the host's native media framework and any codecs available to it, is used. (This will also provide increased performance, as generally the native media frameworks are optimized to use the host's graphics system in the most efficient manner.) Opera may end up having to shell out for an H.264 license.

If Mozilla doesn't provide a way for H.264 content to be played back in Firefox, then Chrome will completely replace Firefox.

(I'll note, BTW, that I'm an Opera user.)

More likely, though, video sites will consider the Firefox userbase as too large to lose, and <video> won't take hold. Flash will remain as the dominant web video format, and every browser will suffer from its goatse-sized security holes.

Yup, the same pragmatism that led to Fash' dominance, and IE-only sites.

The only way to knock H.264 off is to get some very rich company to sue MPEG LA on Xiph or On2's behalf, for some infringement that H.264 commits on VP3/Theora.

Otherwise, H.264 has way too much momentum.

Like I said, I think what's actually going to happen is that this will kill <video> - and ultimately, Flash is going to win this one. (And, by extension, H.264 that way.)

Yup, the same pragmatism that led to Fash' dominance, and IE-only sites. "

Speaking as someone who worked primarily as a "web monkey" during the time periods when those technologies became dominant, I can tell you it was a little more complex than that.

In reality there are several factors that severely limit the choices that web developers have on those matters - things like the overall OS/browser/web technology landscape, the expectations of end users, the demands of clients, etc.

Those are foremost concerns for every web developer I've met or worked with (with the exceptions of hobbyists who just don't have any clients to worry about). A web developer can advocate open standards until he's blue in the face - but there's not much he can realistically do if the client says "no thanks, we'll take cheaper, de-facto standard solution please."

For some of us, it's not that we "support the trolls", as much as we sometimes feel that your analysis of the situations overlooks a handful of legitimate points that the h.264 camp may make -- or ignores certain inconvenient facts about the current web landscape. Really believing that Theora is a good thing shouldn't preclude living in the real world or treating those who disagree with you like they aren't complete idiots.

FWIW, I'd also number myself among the Theora partisans. But I'm willing to see the situation for what it is. There are a lot of deep-pocketed entities arrayed against us, making a win for Team Open here fairly unlikely. It's unfortunate, but that's the way it is, and admitting it doesn't somehow make me an h.264 troll.

Edited 2010-03-25 23:18 UTC

I don't claim at all that you are a troll, not by any means. MPEG LA are the trolls, the person quoted in the original article (John Gruber) is the troll. My point is that nothing is gained about being modest about where Theora is at. Right now, for the purpose of video for the web, Theora can give a better result than h.264.

I know that is not the conventional view, but the conventional view is wrong because it is out of date.

We should be shouting this from the rooftops, not being shy about it.

Edited 2010-03-25 23:35 UTC

Being courteous, fair or realistic are not weaknesses. "I am right and I am noble" is not an excuse to ignore everything your opponents say, and to constantly turn any and every discussion into blind advocacy of your position. You'd probably get a lot better reception, and better results, if you didn't treat the people who disagree with you like they where either complete idiots or traitors to civilization. Or turn a blind eye to the weaknesses in your own argument.

What is good for the goose is good for the gander.

Proponents of the h.264 codec have absolutely no qualms at all about loudly claiming that Theora is rubbish, disparaging it at every opportunity, making bizzaro-world claims that using Theora makes one more susceptible to patent attack (as in the original article, which is clearly the exact opposite of reality), and utterly ignoring any evidence to the contrary.

Why should FOSS proponents be held to a different standard?

Edited 2010-03-26 00:08 UTC

You are sadly very much out of date.

http://jilion.com/sublime/video

Check it out ... download it and play it. It is high quality, Resolution 1280 x 544, 25 fps, two minutes long and less than 20Mbytes.

Edited 2010-03-25 14:03 UTC

I did check it out. The video is all static or nearly-static scenes. Not the sort of thing any codec has problems with.

Show me something where you can actually *see* the artifacts in both videos, like this: http://www.youtube.com/watch?v=q8t7iSGAKik

(Sadly predictable answer: "You shouldn't be using bitrates that low")

This is why it gets just 20Mbyte filesize for two minutes worth of video.

The point is, I have NEVER seen any h.264 that gets anywhere near that low.



Wilco.

http://people.xiph.org/~greg/video/ytcompare/

You can find there three low-bitrate video frame images from youtube h.264 (499kbit), Thusnelda (486kbit) and ptalarbvorm (376kbit), all of virtually-indistinguishable quality.

In fact I just did my own test.

Source video: high bitrate 1080p h.264: http://www.apple.com/quicktime/guide/hd/artbeats_red.html

This was shrunk to 640x480 using ffmpeg (this also messes up the aspect ratio but I don't care about that here):

ffmpeg -i artbeats_red_m1080p.mov -f yuv4mpegpipe -s 640x480 test.y4m

Then encoded using libtheora 1.1.0 (Thusnelda):

./encoder_example -V 300 --two-pass -o thesnulda.ogv test.y4m

And using x264:

x264 -B 300 -p 1 -o h264.mkv test.y4m
x264 -B 300 -p 2 -o h264.mkv test.y4m

So that both encodes were two-pass 300 kb/s. The final sizes are:

h264.mkv: 4842293 bytes
thusnelda.ogv: 4869379 bytes

So Theora is slightly bigger. Despite this, the difference in videos is *MASSIVE* and *OBVIOUS* at this low bitrate (which DOES GET USED!)

I applaud xiph's efforts and I think Theora is great, but it is just a bare-faced lie to say it is as good as H.264.

Here are the resulting videos to judge for yourself:

http://concentriclivers.com/misc/thusnelda.ogv
http://concentriclivers.com/misc/h264.mkv

NB: Both encoders seem to take a while to get into the swing of things, so start judging when you see the bulls.

There you go, h.264 supporter more than prepared to call other people liars. Why should we be polite in the face of that?

Anyway, I have already demonstrated that Theora can get better results. Another poster (WereCatf) has obtained good results using ffmpeg. It is, however, quite easy to get a very disappointing result using ffmpeg from the commandline, one common error is to mistakenly constrain the encoder to using very small buffers.

So why don't you take your high-quality source (Source video: high bitrate 1080p h.264) and try it again using the firefogg transcoder or the Miro Video Converter:

http://openvideoalliance.org/2010/03/miro-video-converter-released-...


That should make it easier for you to get a high quality result for Theora, and avoid the need to again embarrass yourself through falsely accusing others of lying.

Edited 2010-03-26 01:44 UTC

a) I didn't use ffmpeg for either encode.
b) If that really is the problem it is *their* problem. I just used the default settings which is what everyone else will do. If you want me to tweak the settings for Theora, then I also have to tweak the settings for x264 which will presumably make that better too.

But I can see that you've made up your mind.

a) I didn't use ffmpeg for either encode.
b) If that really is the problem it is *their* problem. I just used the default settings which is what everyone else will do. If you want me to tweak the settings for Theora, then I also have to tweak the settings for x264 which will presumably make that better too.

But I can see that you've made up your mind. "

Get real.

It is always possible to make a bad video file out of even the best codec, for whatever reasons (intentional or not), but it is NOT possible to make a good video file out of a bad codec.

The fact that other people have made very decent, competitive bitrate/quality Theora video files means that it IS possible to do so.

I did it without any trouble whatsoever. I simply used one of the methods recommended here:
http://openvideoalliance.org/2010/03/lets-get-video-on-wikipedia/?l...

http://videoonwikipedia.org/howto.html
(step 2 of the howto in the above link, "Convert your video to an open format")

I did it first try. It was easy. I used the default settings (except that I enabled two-passes) in Firefogg.

The fact that you failed just means that you failed.

Edited 2010-03-26 09:56 UTC

He never called you a liar, and was in fact fairly respectable. This is a time when you're being an ass to someone who's being courteous to you -- which makes even less sense than being an ass to someone who's been rude to you.

\meme{Science does not work that way.} Repeatability (and verifiability and generality) of results is (are) an essential component of the scientific process. Just because you've set up one test where Theora gets better quality does not settle the issue: we need repeatable and generalizable tests, ideally tests with multiple trials and large ranges of inputs. It would be disingenuous to claim that Theora has been shown to be conclusively better or worse than h.264 until you've met that requirement.

But failing that, we should at least consider everyone's tests in aggregate. Accusing someone of "calling you a liar" because they present experimental evidence that doesn't line up with yours does little more than demonstrate how deeply and comically you misunderstand the scientific process.

Not that scientific methodology really has any bearing here, I guess.

Edited 2010-03-26 19:02 UTC

He never called you a liar, and was in fact fairly respectable. This is a time when you're being an ass to someone who's being courteous to you -- which makes even less sense than being an ass to someone who's been rude to you. "

Oh, yes, he did.

Here is his quote:


This was in response to a post of mine where I showed direct, tangible evidence of Theora being as good as h.264.

He was rude, not I. I merely called him out on it.

You really need to get rid of your bias, and look at what was actually said.

Edited 2010-03-27 11:27 UTC

So Theora is slightly bigger. Despite this, the difference in videos is *MASSIVE* and *OBVIOUS* at this low bitrate (which DOES GET USED!)

While I can confirm that at high bitrates Theora produces really good results I can ALSO confirm that at this low bitrate it really totally blows: I just did a similar test of my own as the previous one, but I used 300k bitrate and the results were just devastating. Theora video was not only blocky, it was really unsharp.

I honestly didn't expect it to perform this horribly.

Funny, there are results from people involved with Theora where Theora gets better and better compared with h264 when you go to lower bitrates.

http://people.xiph.org/~greg/video/ytcompare/comparison.html

Funny, there are results from people involved with Theora where Theora gets better and better compared with h264 when you go to lower bitrates.

That may be if they use non-default settings and know what settings to tweak? Maybe they even tweaked Theora settings while left H.264 settings at default?

I don't know, I can only say that using default settings on both and with a low bitrate Theora performed significantly worse than H.264 and is not worth using :/

You say that this bitrate is common yet this clip, encoded with what many regard as the the best encoder available (x264), looks terrible. Far below even the famously bad Youtube quality in many places, which makes sense as a quick google claims they use roughly this bitrate for videos 1/4 the size.

If this bitrate *is* used commonly for this size of file without such obviously poor results then perhaps the clip is unrepresentative and contains many difficult to encode scenes (snow, smoke, waterfalls, etc.)?

So it would be a more realistic test if you either used a more typical, easier to compress clip or you gave this difficult clip more bitrate, but if you did that the differences between the two encoders would fade away.

So all your test shows is that if you're going to encode at a rate that makes even x264 look bad, then Theora will look worse. I can see how technically that's a victory but it's not particularly satisfying is it?

All encoders fall off a cliff at some point as you reduce bitrate, pushing two competing encoders right to the edge and then publishing the results as if typical of the one that slips first is either a honest mistake or a cheap trick. Look at any graph of encoder quality and you'll see them all rise nearly vertically, bend then rise at a much reduced rate not much above horizontal. A small change in bitrate can have vastly different effects at different points in the curve.

It would be like reducing the memory available to two competing programs until one started swapping. It might prove that one uses more memory than the other but the resulting slowdowns would in no way be representative of the impact on normal usage.

Edited 2010-03-26 13:15 UTC

Well, the mp3 patent holders were smart enough to change the terms to keep mp3 viable for most people. So far, the MPEG-LA has not done so permanently and even then the temporary web video terms aren't anywhere near as good for individuals as mp3. At least I can't be sued for encoding a file to mp3, even though I typically use Vorbis anyway.
Also, there was never an audio war. Quite a few hardware devices implemented Vorbis support alongside mp3 and there wasn't any hell raised. As for software, audio is typically passed off to your audio player of choice so it didn't matter what format the audio was in so long as it was in either mp3 or vorbis. Aside from iTunes and WMP, I can't think of any current software player that doesn't support Vorbis out of the box. In that way, the situation was a lot more gracefully resolved. Those who want mp3 use mp3, those who want vorbis use vorbis, and no one's screaming at the other side that they must adopt one format over the other. There's no real conflict on the audio side of things and even when there was it didn't last long thanks to Fraunhofer and others getting smart, and thanks to LAME kicking their ass when it came to quality.

Clearly the solution is to use plugins and host them in software-patent-free countries.

No! Clearly the solution is to just burn every pixel horny monkey at the stake and tell MPEGLA to shove it.

(God, why did you make stupidity painless...)

A little less quality and no more corporate taxes, to me, are preferable to your patented, rights limited, pay till you drop, a few pixels for your soul, nightmare world.

The way to get rid of software patents is to make sure there is no money to be made with it. Hosting infringing plugins in software-patent-free countries just puts unnecessary shackles on your fellow humans in software-patent-infested countries.

If there is no dependence on encumbered, costly codecs, there is no need to move into dubious grey areas by hosting in favorable jurisdictions. Everybody wins!

No, not true, we still need to burn Timmmm at the stake. :-P

I haven't found the quality difference to be noticeable, at least with settings that are typically used for video on the web (dimensions of 640x480 / 640x360 or smaller, bitrates less than 1,000kbps, etc).

That said, I'd still be interested in seeing the results of some basic double-blind testing. I'm surprised that no one has done that yet (to my knowledge), as the results would settle the image quality debate fairly effectively.

There are objective ways to measure the SNR for a given video, aren't there? Couldn't we at least theoretically do something like render 600 static images, save them in a lossless format, assemble them into a movie with an h.264 encoder and a Theora encoder, then measure the SNR's?

(I think SNR is the term I want... if I'm using it incorrectly, someone tell me. )

PSNR. Peak signal-to-noise-ratio.

http://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio

H.264 is a couple of db better on this measurement than Thusnelda for the same filesize. H.264 was quite a long way ahead of Theora 1.0 on this measurement.

http://web.mit.edu/xiphmont/Public/theora/demo7.html

Ptalarbvorm, the current still-experimental version of Theora, is at least as big an improvement over Thusnelda as Thusnelda was over Theora 1.0.

It's worth noting that Ptalabvorm is *intentionally* going to reduce PNSR substantially, by introducing psychovisual techniques that look better to humans, but do worse in simple numerical comparisons like PNSR (which as Monty says in that link should probably never be used to compare different codecs as there are too many variables. It can be useful for comparing small improvements in a single codec though.).

SSIM is a more advanced measurement technique which shows substantial improvements in 1.2 even as, indeed because, PNSR is reduced.

There are, but (IMO) the results are mainly useful/interesting from an academic standpoint - they don't tell you much about which codec people will prefer in real-world situations, since that involves a fair amount of subjective perception.

More true than you know since, according to an Ars article I read a month ago, H.265 will be relaxing the frame-perfection requirement to focus more on fooling the human perception of motion.

True. Precisely.

The world is thankfully a bit more of a wakeup to this by now.

What is interesting, however, when you see claims of this nature, is that you can universally make the observation that the party making the vague claims is in fact the fearful one.

It's just entertaining to me. I see claims like that, "oh, no, that product we compete against might be eaten by vague, hypothetical patent-monsters in the unspecified future, you'd better use our product, just to be safe," and my reaction is pretty much, "I thought we as an industry had grown out of this year ago?" Guess I'm naive for expecting that to ever happen.

The Theora decoder may use patented technology too, so may just as well be the target of patent litigation. Patents do not just cover encoding.

Edited 2010-03-25 22:48 UTC

True, but since the decoder has not changed, this has been true for many years. No patent claims have emerged. Remember, Theora is an older format. The patent trolls making these vague no-substance threats against Theora have younger technology. Sorry, but in patent-world, that is a "close, but no cigar" scenario.

I know - I'm not saying the decoder IS liable (remember, I just wrote this article to dispell that?); I just wanted to point out to Adam that patents don't magically stop at encoders.

Yes but Google isn't much of a target since they aren't selling a product that uses Theora which would severely limit any award settlement. Google would most likely just remove the decoder and settle out of court.

And as I said earlier if anyone were to sue it would be MPEG-LA since they are the largest video codec patent holder. If Google dumped h.264 for Theora this would be a possibility but we know that won't happen. Theora really didn't have much of a chance when Google and Apple joined up with the MPEG-LA group.

The harsh reality is that Google has wanted to kill off Theora from the beginning. They knew they could include it in their browser while working against it at the same time. Theora was never held back by patents, it was held back by Google refusing to use it in YouTube.

Since Theora is the older codec, any patent decision about common patented technology is more likely to go the other way ... older "inventions" trump newer ones which use the same methods.

So, do you think that MPEG LA would most likely just remove the decoder and settle out of court? What about all the fees they have been unfairly extracting to that point?

Neither side wants a patent war but MPEG-LA would be willing to settle for millions if Theora actually had a case. The Theora group isn't in the movie business so a compromise would likely be reached. MPEG-LA isn't pushing HTML5 and would be content with having their codec stay inside plug-ins.