New Open Video Codec From Xiph/On2

xercist writes: "Xiph.org, the bringers of the mighty Vorbis codec, have done it again. The patents on On2's VP3 video codec have been effectively neutered, and it is being released under the BSD license for all to enjoy. The combination of VP3 video and Vorbis audio (in an OGG bitstream, of course) will be called Theora, and will soon take over the world. The ETA to a 1.0 release is approximately one year. You can also read an interview with Emmett Plant (Xiph CEO) here. The official press release will be up tomorrow, so don't complain about lack of mention on xiph.org just yet."

Comments

very nice but can it overtake DivX?

[Savatte]

It's an uphill battle, that's for sure, and the DivX team is sure to come up with new features in a year. But does this open codec have a chance to overtake DivX? This could be another mp3/ogg debate, but more choices in video codecs is always better.

Re:very nice but can it overtake DivX?

[Goonie]

Is it possible under the BSD license they mentioned that they can keep the algorithms secret?

In a word, no. The Xfree project's license is a good example of a modern "BSD license".

Note that you can incorporate portions of the code into your proprietary product, which doesn't have to be released under the same license, and that you don't have to provide source. As far as the recipient of the software is concerned (who might go on to use sections of it in their own proprietary products), it is more liberal than the GPL.

Don't get confused, though, the developers, Xiph, still provide full source and thus the algorithms are completely public.

The reason the BSD license is chosen for this project is presumably the same reason the Vorbis libraries are BSD-licensed - so that VP3 support can be incorporated into proprietary software, which as I have posted elsewhere in this discussion is IMHO a good thing.

Re:very nice but can it overtake DivX?

[xeromist]

I think it is very possible. The reason ogg really hasn't taken a significant portion of the market away from mp3 is that mp3 is so well established. It is so quick and easy to make an mp3. It's very easy to come by software for making mp3s and you can rip and encode an entire album in less time than it takes to play it. Anyone can do it.

By comparison, DivX is widely used, but _not_ established. Here's why:
There are far fewer people making DivX movies than mp3s;
The software is slightly more difficult to come by;
Ripping takes longer because most DVD drives are slower than CD;
The raw video takes a lot of disk space;
Encoding takes a VERY long time compared to mp3.

Overall this means that those making DivX tend to be a few savvy users and it will be easy for those users to switch to a new standard. So that being the case it will only take a few people changing over to a new standard to affect a change. Of course these people will have to see that Xiph/On2's way of doing things is better, but if Xiph/On2 can prove that, I'd say they have a very good chance.

VP3 and quality

[Vardamir]

Does anyone know how vp3 compares to other codecs that are already here or emerging, namely MPEG3 and DivX? Ogg is obviously making headway in the technical area of being the best lossy audio codec, and if vp3 can do the same, then hopefully this will give Free Software an edge in the media areas.

One problem here though; I suggest someone adapt the VP3 code to a GPL license, ortherwise Microsoft, Apple, or any other company could simply take VP3 and make it Free Software's worst enemy by not releasing specs on the derivative audio codec. Observe: we are just now beginning to see Sorenson codecs that are open source.

Re:VP3 and quality

[Webmonger]

Usually, I'm a GPL supporter, but I can understand the reasoning behind using BSD here.

They're trying to establish a new video standard, which isn't easy. So they want to encourage the maximum number of developers to participate. A BSD license will do that.

It's questionable whether someone would spend the effort to take the codec, improve it significantly, and make it proprietary. In any case, such a proprietary version would probably not catch on.

Remember, there are no effective patents in this case, so the Sorenson problem would seem not to apply.

And Xiph now has a track record of actually maintaining their code, making it better, not pulling funny tricks.

Oh, I (or you) can make this new code GPL today. Just download it, change the license at the top, and post it. But what would be the point of that? There'd still be the BSD version, and it would be better-maintained.

Re:VP3 and quality

[mdarksbane]

I've used VP3 quite a bit, since DivX hasn't released a mac encoder yet. In my tests (recording on a Tibook G4 667) it encodes more slowly than Sorenson 3 or Apple's Mpeg4 codec, but the quality is very comparable, even better in high-motion scenes.

Mpeg4 and Sorenson 3, even at bit rates nearly half those of an Mpeg-2 DVD stream, still produce pixellation artifacts in very high-motion scenes. VP3, at any bitrate over 30Kps, does not. While it is not as sharp as the other codecs at high bit rates, I found it to be very superior at lower ones.

My only complaint is that, for some reason, any movie encoded at full size (640 by 480) would, not matter what the bitrate, barely play back at all. even though full 30 fps video plays back without difficulty on my G4. It wasn't a case of a few dropped frames, but a total stall down to 4 fps.

Anyway, for high-quality, low-bitrate video, it's the best codec I've had the opportunity to use personally in terms of quality and playback/ kbps. I'm sure that the newest DivX surpasses it, but I won't be able to play around with that codec until they release a mac encoder. VP3's quality is comparable to the DivX movies I have downloaded, though.

With some development, it could be a very competitive offering.

Theora?

[SpamJunkie]

Xiph needs to realise they don't know how to name things. A good name is part of marketting. It doesn't matter if you and all your geek buddies can talk about Ogg Vorbis without feeling akward if all the other kids think it sounds like Klingon. Is it any surprise DivX is so widely used when the name was basically stolen off a product that had commercial marketting?

VP3 is a great name. Most people can even imagine what it stands for - video mp3 - which they would be very comfortable talking about. Why change it?

Re:Theora?

[xercist]

Like all Xiph projects such as Vorbis or Tarkin, Theora is named after a fictional character. Theora Jones was the name of Edison Carter's 'controller' on the television series Max Headroom. She was played by Amanda Pays.

Porn films

[rodgerd]

We all know that porn site's choice of encoders dictates the success of media formats!

Re:Porn films

[brooks_talley]

Porn sites' choices of video codecs are based more on installed base and ease of use than technical merit or licensing costs (remember, a decent porn site is profitable and can afford to spend money to make money). Far better to pay for a solution that your clients will be able to use than try to get away with a free solution that will do nothing but cause support headaches and (the bane of adult sites) chargebacks from dissatisfied customers.

Anything that involves a download and/or installation is bad news. People don't trust downloads from porn sites (though, to the best of my knowledge, there's never been a case where a porn site used a download for nefarious purposes). Beyond that, as we all know, some percentage of installations fail for one reason or another. If you have to supply even a link to a third party download, people will blame you if it doesn't work.

For these reasons, Windows Media is clearly winning the battle for porn site content, measured by new clips produced.

Real probably still has an overall lead in clips available for legacy reasons, but nobody is producing new porn in Real format. It comes back to support and chargebacks: Real tries so hard to trick people into the "free 14 day trial, $9.95/month" player that it generates no end of customer support headaches ("You said if I paid $29.95 a month I'd get free video... and now I have to pay another $9.95 a month for the player! You bastard porn sites are all the same! Lying, cheating scum! Why can't you operate like normal, ethical businesses?").

Not that I'd know anything about that industry.

Cheers
-b

excellent!

[Bobzibub]

What the headline implies is that On2 is not helping out. This is incorrect.

"On2 will sponsor work done by the Xiph.org Foundation to combine those projects, plus On2's already open-sourced VpVision personal video recorder, into a product they believe will provide serious, free competition to the increasingly expensive MPEG-4 royalty fees." --from NewsForge

Hats off to On2!
-b


PS. Currently listening to Oggs and ripping them too. tres nice.

Just a note...

[jmd!]

Before anyone praises "On2 Technologies" too much, you should be aware of their product line. VP3 is their old leftovers. They've since improved it and release VP4, and recently made more improvments and released the VP5 codec. They've been giving VP3 away for some time, as more of a demo of what their newer technologies can do better.

While it's better then nothing, (we certainly need an open video codec), but On2 isn't exactly fully on the Open Source wagon.

Just a few facts for those who didn't and won't bother to read the background and articles.

Re:Just a note...

[benwaggoner]

Not entirely true.

VP4 was a set top box codec. It was never released in any form for computer-based playback.

VP5 hasn't yet been released. I don't know what On2's plans are with it.

Pass the crack pipe

[MisterBlister]

This will take over the world?

Sort of how OGG took over the MP3 world? Not likely. Its nice to have an optional patent-unencumbered method for video & audio sharing, but anyone who thinks this will be anything more than a tiny niche product living in the shadow of MPEG4/WMV really needs to get out of his mom's basement more.

Unlikely

[Goonie]

Microsoft and Apple already have plenty of proprietary video and audio codecs, some of which are de facto standards, so if they incorporate an incompatible hack of "VP3" that's bad luck, but doesn't actually makes things any worse.

If it's GPL'd, the above isn't possible. However, it's also much more difficult to incorporate unhacked VP3 support into their players and encoders, because they have to write their own code rather than just using the available library. That is bad, because we (the free software using community) *want* them to add VP3 support to their proprietary players. Let me repeat that - VP3 support in proprietary software is a good thing.

Why, you may ask? Because if it is available in the software that everybody uses (which, sadly, at this point is still proprietary software), it might become a de facto standard, become the standard format web video clips, for instance, are encoded in, and thus free software users are no longer second-class citizens when it comes to video codecs. Additionally, it makes the ultimate migration of Windows users over to free desktops that much easier.

I believe even RMS has agreed, on occasion, that the BSD license can be appropriate if it helps free file formats become the standard.

Re:How exactly...

[Servo5678]

How exactly do you "neuter" a patent?

With snippy scissors and a cold, cold heart.

"Hold still, little patent, this will only hurt for a second..."

another xiph open standard?

[joe_bruin]

they havn't gotten around to polishing vorbis yet, where do they get the time to work on a video codec? will this be the same maneuver as the ogg format?

"look, it's free and open. well, unless you want the specs, which don't exist, so you have to use our source. want an integerized implementation (for your rio)? well, since you don't have the specs, we'll be glad to sell you one."

this is why ogg is not in hardware yet. this is why there arn't alternate implementations (LAME ogg, anyone?).

i'd love to see a free, open video codec. mpeg royalties suck ass. but my patience with xiph is running short.

eat more chicken!

Re:But what about more than 2-channel audio?

[uhoreg]

Read the last question of the FAQ. "Vorbis does currently support greater than two channels; the default multichannel mapping in the 1.0 release supports up to 255 simultaneous channels." (Mmm. 255 channels.)

Oh, great.

[WasterDave]

So I take it they gave up on Tarkin then?

FWIW there's a (getting old) codec comparison on Doom9 (http://www.doom9.org/codecs2.htm#test1). VP3 comes out *really* badly.

Dave

Re:Oh, great.

[Forkenhoppen]

Is that a codec limitation, though, or is it an encoder problem? This, of course, is the big question, because if it's just the encoder, that can always be improved.

LGPL bad for embedded devices

[Straker Skunk]

Embedded devices often use the equivalent of statically-linked binaries, i.e. ROM images. The LGPL requires that you allow the user to swap out the LGPL'ed part [of a larger proprietary solution] as desired. If you distribute software which is inseparable (to the user) from LGPL'ed code, then the entirety of that software must itself be under the (L)GPL---which would allow any user with a binary to ask for source to your stuff.

The BSD/X11 licenses, since they allow outright proprietarization of their covered code, naturally don't present this difficulty. (FYI, the SDL is well-served by the LGPL as---to my knowledge---the project is not targetting the kind of embedded systems where this would come up.)

How to help make it popular...

[Dr.Dubious DDQ]

Finally! A formal statement that specs will finally be written for video+vorbis in .ogg

From what little I've seen, VP3 is, overall, not as "good" as the various MPEG4 variants out now, but is a little better (in terms of quality and lack of "artifacts") than the "windows media" implementation [at least, from the one review I looked through].

The important thing from my perspective is that VP3/Vorbis in Ogg will give us a completely "free" way to offer videos...which brings me to my point.

There ARE some "public domain" videos out there. Not just obscure "indy" things but actual commercial movies, cartoons, shows, and so on that matured into the public domain when their owner didn't renew them (back when that was required).

There's a whole mess of them available on LSVideo (which appears to be undergoing a redesign, but offered and will apparently continue to offer a wide variety of public-domain [i.e. you can legally make copies for all of your friends if you want] videos) and RetroFilms. Retrofilms even offers a number of Disney (!) cartoons that slipped through their iron grasp into mature public-domain works. MOST of them are rather old, but many are well known (Metropolis [not the new Anime', the classic silent film], for example, and the classic "Nosferatu"...and, I believe, the insipid [in my opinion] but well known "It's a Wonderful Life".)And, of course, there's a whole mess of interesting and/or bizarre and/or educational things in the Prelinger Archives Movies Online.

So....as soon as encoding software becomes available [I suspect ffmpeg and/or MPlayer will be set up to handle it pretty quickly after the initial source code and specs become available, if their recent development speed is any indication] I plan to go through the surprising number of videos that I own that turn out to be Public Domain, encode them into "Theora"-type files, and try making them available peer-to-peer.

At the very least, there are bound to be enough "oddball" videos available in the public domain that making them available in this format, combined with the fact that neither the "content" nor the file format, nor the video codec, nor the audio codec will be legally prohibited from distribution, they could easily become encountered often enough to promote the format to the point that, though it may never actually overtake proprietary formats, it'll pretty much "need" to be supported by any piece of multimedia software and playback unit that intends to bill itself as handling a lot of different formats...

I yearn for the day when my DVD playback unit can handle "Theora" videos and "Ogg/Vorbis" sound in addition to the .mp3's it already does...

Get cracking on that spec, Xiph!!!!!

(P.S. - Are there already IRC channels devoted to serving legal, public-domain videos?...)

Bla bla bla No numbers! c'mon people.

[tcc]

I read the discussions here, the interresting subject is the licensing, the FUD subject is how it will take over and so on.

First, before even talking about taking over Divx, please point me to a site that has a codec study, not on the theorical, not on lame useless "tom's hardware" numbers, but on a scientific or scientific-like approach with proper setup and testing.

Nothing like an example to clear things up:

Codec usage, yes it can crunch more than divx for "similar" quality that is "similar" to DVD, which is "similar" to uncompressed video... heh. no, this is totally wrong, what I want to see is uncompressed video, original sequence, then compress Divx with best quality at X bitrate, do the same with that new codec, try at least 3 different kind of video that are abusing codecs differently: High contrast (black and white with sharp edges being the most evil), smooth colors (4:1:1 DV videos can show some serious color bending when you look at it closely, for example) so how is the color bending affecting each codecs?, and finally, motion, low speed, high speed, repeat with the last two example and make all the different mix, and then try different bitrates, then check the results, if they are similar, zoom in, check the pixelization, I don't want just words or lame tests with no numbers, I want a half-decent scientific approach, I wish I had time to make such a thing like I did before with mpeg/mjpg/mpeg4 V1.

Of course we all know that you don't have to be the best to win, there are other factors to consider as well, but in my case, QUALITY will make me switch from a codec to another for my video presentation, so far, I still use DIVX for video-only movies, a good optimized MJPEG codec for software playback of home-made multimedia cd-roms on PC only (divx needs too much cpu juice that I require for other realtime tasks on my video display software for slower machines), and I just wish I would find that perfect codec that wouldn't look like thrash when I display the colors and motion on a plasma screen.

I will jump on anything new if it does the job better than what I have right now, but please, don't just say it's so much better because of [blabla], say actual facts and numbers, it will be very interresting to read.

MPEG-3? (MPEG taxonomy)

[benwaggoner]

There was no MPEG-3. That was the working name for the original high definition MPEG format. However, they decided that they could implement HD with extensions to MPEG-2. Thus, MPEG-2 is used in HDTV, and there is no MPEG-3.

MPEG-4 is the new video/audio/streaming/etcetera standard.

http://mpeg.telecomitalialab.com/standards/mpeg- 4/ mpeg-4.htm

There are no MPEG-5 or MPEG-6

MPEG-7 is a forthcoming media metadata format. It doesn't include video compression technology. You'd still use MPEG-4 codecs within MPEG-7, or even use non MPEG codecs.

(The official link is broken right now)

No MPEG-8 through MPEG-20, at least not yet.

MPEG-21 is a multimedia authoring and delivery format. It's in very early stages, but think more like a competitor to Flash MX, writ large. We're some years from seeing products based on it.

http://mpeg.telecomitalialab.com/standards/mpeg- 21 / peg-21.htm

What file format for Theora?

[benwaggoner]

From reading the announcement, I don't have much idea what file format this is going to use. .ogg as I understand it is really designed as an audio codec/format; I'm not sure how easy it would be to add video samples to it, usefully.

If they're looking to still pick a format, I hope they do QuickTime. QuickTime's file format itself is open and documented, and there are a number of open source projects to implement it. As of QT6, QuickTime itself now has native support for VBR audio encoding, which makes it easier to do an Ogg encoder inside QT (VBR decode has been in there since 4.1).

http://developer.apple.com/techpubs/quicktime/qt de vdocs/PDF/QTFileFormat.pdf

With QuickTime support, one immediately inherits a wide installed base of players, and lots of functionality. For example, real time streaming support is availble via the open source Darwin Streaming Server. The codecs just need to have a native packetizer added.

The nice thing about implementing Theora as a QuickTime file would mean folks would have the option of using high-end QuickTime encoding apps like Cleaner for encoding, and generally letting the files work well in both the closed and open source universes.

VP3 is of course already implemented in QuickTime so doing this would mainly be a matter of finishing the Ogg port as a QuickTime codec:

http://qtcomponents.sourceforge.net/

Re:What file format for Theora?

[ianezz]

.ogg as I understand it is really designed as an audio codec/format

Read better: Ogg is designed to be a generic container for audio/video streams, exactly as Quicktime. The audio codec part is Vorbis, which is shipped inside an Ogg file, hence the full names "Ogg Vorbis".

I don't know why they choose to implement a new container instead of using Quicktime, but probably they have their reasons.

Re:What file format for Theora?

[inquis]

You seem to be confused as to what ogg really is. Blame part of this confusion on Microsoft, who would lead you to believe that because a file has the extention .ogg and plays audio when you run it that it's an audio file.

The ogg format is a stream format, much like the avi format (published by Microsoft) is. Ogg's primary advantage is flexibility: you can embed practically anything into an ogg stream (video, audio, subtitle streams, whatever) that your player can interpret at the other end. Furthermore, the stream is fault-tolerant and should play even if damaged or incomplete.

(The ogg audio file you're thinking of is actually a vorbis audio stream multiplexed into an ogg.)

Probably the best place on the web right now to learn about video codecs and stream formats is doom9. You will also find that in the forums they are doing active research concerning a new stream format called mcf (well, actually an old stream format, I think it was specified before ogg) and a way to burn raw mcf streams to a CD a la VCD.

Doom9 also performs codec comparisons every time a company / individual / group thinks that they've produced the Latest and Greatest video codec. VP3 was obselete many months ago when it was being compared to div3 and even wmv8, both of which are suboptimal when compared head-to-head these days with div5 and xvid.

Another thing to keep in mind is that VP3's patent-unencombered state is a two-edged sword: while you don't have to worry about maybe having to pay MPEG4 license fees, you also don't have a chance in hell of being playable on any next-generation DVD player. Sure, if you're a game manufacturer and you're looking for a video and audio codec to use in in-game movies, you'll probably use an ogg file with a VP3 video stream and a vorbis audio stream; however, if you're producing video of your son's first steps that you want to throw into your grandfather's new DVD player, you'll be doing it in some variety of MPEG4.

-inq

Because it's finally useful

[xiphmont]

Open Source's strength is in making commodity software, the software people actually *need* and *care about*, free and open... part of the infrastructure of democracy if you will. There was no truly free audio before Vorbis, because only about six users in the world had computers that could play decent audio in 1993 (when the work on Ogg began). And by decent, I don't mean FM synth or mono 22kHz.

Similarly, there was no need for Open Source video until now because... no one used computer video in any great numbers until now. Yes, yes, Quicktime has been around for a while. I'm personally not excited about the 80x60 black and white 4 frames per second stuff from 1990, and neither is anyone else.

Real and Apple pushed proprietary standards at a time when it was not at all clear that many people cared about 'multimedia', and those that did could afford to shell out big money, or buy into a single vendor. That time is quickly passing. I use both audio and video on my own boxen every day, and Mom in Ohio is likely only a few years behind me on that front, so it's time to get ready.

(Nor was Quicktime 'first', you've simply forgotten its predecessors. That has nothing to do with my point above.)

Monty
xiph.org