On2 VP7 dual license ?

[DoomNoob]

Hello the world, first post here ;)

I have a very simple question, after reading D9's 2005 codecs comparison it was obvious to me that Theora is very obsolete & Dirac is far too greedy ...

as I am a big Ogg Vorbis fan myself, I was quite disapointed, so I started seeking for a solution ... & I had this stupid idea :

- On2 gave VP3's source-code to Xiph for free
- On2 gave VP7's binary to End-Users for free
- On2 seems to make his money from big compagnies, not from end-users

so why couldn't On2 release VP7/8 as open source but patented technology & use a dual license like Trolltech's QT4 ?

- Open Source & Free & Patented for End Users
- Open Source & Buyable & Patented for On2's Big Compagnies customers

the big advantage of this would be that On2's VP could be added to linux distro by default, giving VP7 all the nix community behing him & making it immediatly a standard like VC-1 is for windows or mpeg-4 for macs ...

Is there anything that would prevent On2 to do so if they were willing to ?
Do they really own the patents of all the technology they use ? Or is there some 3rd party submarine patents (mpeg ?) in their code like there is in MS VC-1 ?

Why couldn't On2 be the Trolltech of linux video & Xiph be its KDE ?

Is it so stupid that I must forget about it immediatly ? ;)

[CruNcher]

You are not the only one dreaming of this to become truth
http://forum.doom9.org/showthread.php?p=691920#post691920
;)

[savage747]

I have a very simple question, after reading D9's 2005 codecs comparison it was obvious to me that Theora is very obsolete & Dirac is far too greedy ...


Well, there are some remarks I'd like to do regarding Theora performance in this test:

Doom9 used single-pass ABR mode with the Theora reference coder - without mentioning this.


All codecs were tested in a 2 pass setup using the settings suggested by the developers where applicable. At this time, Dirac only offers one pass encoding.


I couldn't find a word about

Theora only being single-pass at the time being (main target: streaming)
ABR mode being *buggy* (poor encoding quality in comparison to what ABR could archieve). This is widely acknowledged amongst the Theora developers and Doom9 was informed about this.


No, I don't say that Doom9 should have used a different mode. ABR is the only available mode for Theora yielding more or less predicatable bitrates (although it is buggy) - so there was no other option for DVD backup. It would have been nice to inform the public audience about this to put results into perspective.

Enough ranting here ;)

The main effort when developing Theora was writing a nice and complete spec of the format ( http://www.theora.org/doc/Theora_I_spec.pdf ). This effort was a good success IMO. As you can see Theora was expanded a good deal over what VP3 had to offer (e.g. several quantization matrices per frame on a per-block basis, more color-spaces, flexible codec headers allowing further increase in compression efficiency and whatnot).

Theora's main problems at the time being:

The Theora reference encoder basically is the good'ol VP3 encoder from ca. 2000. The encoder was changed to output the new bitstream format - but nothing was done to increase actual coding efficiency. It doesn't use any of the features the new spec has to offer.
Theora's developers are short on time. There's a new experimental Theora code base written from scratch - but it's just not ready yet (it produces valid bitstreams using all new features - but the encoder's psychovisual model is broken)
If Xiph.org was considered to be something similiar to the MPEG consortium 3rd parties would take the spec and the reference code (which - in case of Theora - is in a better shape than any MPEG reference code) and implement their own, optimized versions. This isn't happening, though, for whatever reason.


IMO we don't need "Theora-II" based on VP7 or whatever but a good implementation of what already is there (nicely documented, cross platform, stable and well engineered).

If On2 would develop a (commercial) Theora encoder - now, that would be nice. They have the video coding expertise, they can still recognize Theora as being their grown-up VP3-baby and therefore surely have what it takes to develop a kick-ass Theora encoder. Plus they could celebrate a media-stunt about developing "open standards" software. I assume they don't have the man-power to pull that off though - and business decisions usually work different anyway ;)

Another possibly nice thing would be a Theora encoder based upon the ffmpeg framework. They have stable, fast and high-quality algorithms for whatever is needed for video-coding and judging from the magnitude of formats that ffmpeg can encode I assume it would be possible to create a Theora encoder on top of that. As the bitstream syntax of Theora is different from anything MPEG that may result in an encoder not violating any MPEG-patents, too (although some algorithms may have to be avoided, IANAL).

And yeah, this is a posting beyond any realities ;)

[bond]

why should on2 have an interest in allowing anyone to make money with vp7 if they can make all the money themselves? ;)

at theora-in-ffmpeg: there is already an own theora decoder, independant of xiph source, in ffmpeg

[DoomNoob]

well nobody is making money out of QT if Trolltech itself doesn't make money of it first ... it must be possible to built a similar license with a video codec ... in no way I am speaking of releasing VP7 code for free ... or lowering On2's incomes ...

about Theora quality ... well I have seen many Theora preview encodes/streams ... be it Fluendo or Honey, IMHO the only thing Theora rivalizes with so far is ... mpeg-1 ... so I don't think D9 comparison is so far from the truth even with a sub-optimal setting ... sad but true :(

[bond]

the current reference theora encoder not that bad qualitywise imho, but its basically only a reference encoder not showing the full capability of the theora format

[DoomNoob]

well to be fully honest, Theora rivalizes with a bad/early mpeg-4 ASP codec ... but in 2006 does it stands any chance against AVC ? Xiph needs something to fill the gap between ASP's mid/low quality & Dirac's high hardware requirement, & it seems to me the saviour can only be On2 ... or we will all be "nero-bobololo-dimkovic-ified" soon ... if you see what I mean ;)

[bond]

read my ogg theora sticky and you will see that theora is in not so few regards better than mpeg-4 asp
i would guess its possible to write a theora encoder that is on par with current asp codecs

[Inventive Software]

So all it requires is a full-featured implementation, then we can make the judgement.

[Doom9]

Doom9 used single-pass ABR mode with the Theora reference coder - without mentioning this.Not true! First there's so that gave me a video bitrate of 581 kbit/s for Theora. which clearly points out that bitrate based mode is being used (and making any other assumption is plain stupid.. you don't calculate bitrates if you're going to use another mode and in a DVD backup scenario with a fixed target size, there can only be bitrate based encoding).
Since Theora has no 2 pass ratecontrol, Also points out there's no two pass encoding.

And what good is the best standard without a good encoder to make use of it? The first AVC encoders delivered subpar quality and were slow. Ateme was the first company to come up with a high quality and fast implementation.. nowadays there are more but there are still a lot of implementations that produce crappy results. For Theora there's but one encoder and it doesn't do good.. and there's no putting that into perspective.. it is what it is. The comparison just happens to look at DVD backups and not what an encoder can do apart from that or could be good at. You said it yourself, the theora reference encoder isn't good..

[savage747]

Not true! First there's which clearly points out that bitrate based mode is being used (and making any other assumption is plain stupid.. you don't calculate bitrates if you're going to use another mode and in a DVD backup scenario with a fixed target size, there can only be bitrate based encoding).
Also points out there's no two pass encoding.


It's clear that a bitrate based mode was used.

"Finally, I was told by people working on Theora that the Ogg overhead is 1.08%, so that gave me a video bitrate of 581 kbit/s for Theora."

Sorry, but I don't see how this indicates the bitrate mode used was single pass. This could just as well point to a broken 2-pass encoding mode.

In the first quarter of the first page:

"All codecs were tested in a 2 pass setup using the settings suggested by the developers where applicable. At this time, Dirac only offers one pass encoding."

A simple "Dirac and Theora" would clear up everything. Just reading that sentence may lead a majority of readers to deduce that *all* codecs (including Theora) but Dirac were tested in 2-pass ABR.

My bad: I indeed missed "Since Theora has no 2 pass ratecontrol, I figured I'd still have a look at the result to see if it makes any use of the almost 20 MB oversize." which is in the last third of the first page. Mea culpa. I stand corrected and I apologize. I still wonder why Dirac got a special mentioning for lack of a 2-pass encoding mode (and Theora not) though.


And what good is the best standard without a good encoder to make use of it? The first AVC encoders delivered subpar quality and were slow. Ateme was the first company to come up with a high quality and fast implementation.. nowadays there are more but there are still a lot of implementations that produce crappy results. For Theora there's but one encoder and it doesn't do good.. and there's no putting that into perspective.. it is what it is. The comparison just happens to look at DVD backups and not what an encoder can do apart from that or could be good at. You said it yourself, the theora reference encoder isn't good..

I agree. I don't question your results.

[Doom9]

Sorry, but I don't see how this indicates the bitrate mode used was single pass. This could just as well point to a broken 2-pass encoding mode.ffmpeg2theora has no two pass mode as you well know ;)

I still wonder why Dirac got a special mentioning for lack of a 2-pass encoding mode (and Theora not) though.That's just an omission, not intentional.

And it's not ABR but VBR.

[PatchWorKs]

Also points out there's no two pass encoding.

Uhm... strange. :confused:

As already said, i succesfully backuped a DVD in 2-pass theora/vorbis (MKV, OGM had seeking problems) file, using ffdshow (http://ffdshow.sourceforge.net/) & FairUse (http://www.fairusewizard.com/).
Quality is decent (it remembers me the early DivX days), anyway the problem is different, in my opinion.
Why to ask On2 to open source their code when we already have XviD ? And what about Snow ?
There are many great open source developers that needs to reach a goal, i think.
Open source a codec ? What about Rududu (http://forum.doom9.org/showthread.php?t=52325) ?

Just my (stupid) opinions, anyway. :rolleyes:

[savage747]

ffmpeg2theora has no two pass mode as you well know ;)

You didn't write that test for me did you? :)


That's just an omission, not intentional.


Oh, sorry, didn't want to imply you'd omit information intentionally


And it's not ABR but VBR.

I usually refer to VBR when thinking of fixed quality and to ABR when quality levels are adjusted to match a certain bitrate. That's at least how things are named in Vorbis-land ;)

[savage747]

Why to ask On2 to open source their code when we already have XviD ? And what about Snow ?

Well, the main motivation for codecs like Theora is to avoid license fees for all sort of usages (even commercial). XviD and similiar codecs use methods covered by MPEG-4 patents and thus aren't truely "free" (cannot be shipped with regular Linux distros and other commercial products without paying for patent licenses).

AFAIK Snow *may* be covered by patents as well.

[bond]

As already said, i succesfully backuped a DVD in 2-pass theora/vorbis (MKV, OGM had seeking problems) file, using ffdshow (http://ffdshow.sourceforge.net/) & FairUse (http://www.fairusewizard.com/).
Quality is decent (it remembers me the early DivX days)ffdshow has added 2pass to the reference theora encoder

[Sirber]

would help if ffmpeg2theora could be updated for AVS input and 2 pass...

[PatchWorKs]

Well, the main motivation for codecs like Theora is to avoid license fees for all sort of usages (even commercial). XviD and similiar codecs use methods covered by MPEG-4 patents and thus aren't truely "free" (cannot be shipped with regular Linux distros and other commercial products without paying for patent licenses).
Err... well i mean: XviD (and other open source codecs) developers should take this in more consideration, i think. :devil: