Interview With BBC Dirac Developer Thomas Davis

arclightfire writes "The subject of the BBC video codec Dirac has been here before, but we've managed to get an interview with Thomas Davies, Senior R&D Engineer at the BBC who devised the Dirac algorithm. Interesting to note that the codec should be with Mplayer soon; "As far as players go, we'll be submitting a patch to Mplayer to allow it to play Dirac pretty soon." And info about the tech developments in Dirac; "I used tried and techniques, like wavelets, which weren't in standards at the time, and tried to develop them. And that's what we'll continue to do as the algorithm develops. So we've tried to build on some pretty well-understood technology, and also tried to do some new things with it. We're patenting the new stuff, quite a bit of which hasn't got into the software yet. The license means that these patents are licensed for free within the Dirac software.""

Dirac's not the only new codec in the running

[Frac]

h.264 (now formally known as AVC, the video-equivalent of AAC) is maturing. Well, at least Ahead's Nero Digital [nerodigital.com] implementation of h.264 is maturing. The marketing spin has it called Nero Digital, but it's fully h.264 compliant. And the results are impressive [doom9.org] - beta testing shows comparable quality of XviD at half the bitrate.

Why bother?

[JPyObjC Dude]

H.264/AVC is open and extremely powerful. Why bother with another protocol?

From Apple Tiger h.264 page [apple.com]:
Not only is H.264/AVC very efficient, providing extremely high quality in smaller files, but H.264/AVC is also scalable, producing video for everything from 3G for mobile phones to High Definition (HD). H.264/AVC can create great-looking 3G mobile content at 50-160 Kbps, excellent Standard Definition (SD) video at 800-1500 Kbps, beautiful HD video (1280x720, 24p) at 5-7 Mbps and full HD video (1920x1080, 24p) at 7-9 Mbps. So at today's SD DVD data rates, H.264/AVC can deliver full HD. In fact, H.264/AVC was ratified by the DVD Forum for inclusion in the next-generation HD DVD format.

Re:For Linux?

[lphuberdeau]

RTFA

They plan on releasing open source and submit a patch for mplayer themselves.

Re:Dirac's not the only new codec in the running

[Anonymous Coward]

Indeed, realtime compression using H.264 at HD resolutions was demostrated at IBC recently, it's a mofo in terms of hardware but I can rememeber when MPEG2 encoders needed to be chilled, "you can comprese a 270Mbps raw stream down to just 8-9Mbps" they cried in amazement, lol, 8Mbps, if only.

H.264 is pricey even more than MS's WMV 9

[riversky]

The license costs on H.264 are the most expensive on the planet (MPEG LA controls it with a hell of a lot of patented stuff from various companies). This is why it isn't being adopted rapidly. Even MS undercut it with cheaper licensing on Windows Media. This will kill it, a very nice technology. I am all for open source codecs. Perhaps BBC is on to something.

no. Here's why.

[don.g]

mencoder at present (AFAIK) only outputs AVI, a format which doesn't cope well with VBR.

Until it supports the ogg bitstream format, you're not likely to see Vorbis audio support, let alone Theora.

Re:They could at least write it with ResEdit!

[Anonymous Coward]

[Posting anonymously because I know some of the people involved]

Thank you to parent and grandparent. I've been following the Adams Platform/MWC saga for some time, but hadn't seen the most recent ASX document...

Adam Clark and his father are total fruit loops. Quite a few big names (who should have had better-tuned bullshit detectors given their positions) got suckered by them but the smarter ones got out earlier when Adam refused to let even his company's own board independently review the technology.

Frankly from what I hear about Adam's behaviour, they should have been asking pretty hard questions about his mental stability straight away...

Now the Fraud Squad are investigating I'm really keen to see how he pulled off the scam! I know people who probably know how it was done, but I've so far had to play an annoying sort of "warmer/colder" guessing game with them as they're still bound by legal agreements.

Re:"legal in hungary" does not change anything

[ncc74656]

Further- just because [MPlayer]'s legal in Hungary doesn't mean it is legal anywhere else- which is why mplayer isn't distributed with, for example, Debian. I don't believe it is, in fact, distributed with any major Linux distribution.

There is a Gentoo ebuild for it. The MPlayer files (source and DLL packs) aren't hosted by any of the Gentoo mirrors, but are retrieved from the MPlayer website. The same applies to most software for which ebuilds exist, though. Source for GNU programs gets pulled from GNU mirrors, source for SourceForge-hosted projects gets pulled from SourceForge mirrors, etc. Whether you could call this state of affairs "distribution" is open to debate, I suppose. While what Gentoo provides directly is in a sense little more than instructions and a patchfile, installing MPlayer is no more difficult than installing Emacs or KDE.

(It's worth mentioning that system-specific optimizations (-march=athlon-xp, or whatever) are disabled in the MPlayer ebuild, which goes against one of the main reasons people use Gentoo. With that said, it's behaved reasonably well IME.)

Re:mplayer is bloated and going nowhere

[0x0d0a]

Who cares about mplayer support?

Those of us who prefer our keyboard to our mouse. And other folks.

It's bloated. On both linux and MacOS X, it consumes considerable CPU resources- and that's with the fancy interpolation it supposedly does turned off. My Powerbook G4, for example- mplayer consumes about 60% CPU, enough to bake my lap and turn on the fan after a while. VLC, on the other hand- needs about 20%, keeping my lap happy.

You misconfigured it. This is, admittedly, not hard to do, but mplayer is the fastest of movie players if used correctly.

Try using the following command: mplayer -vo xv -fs=yes moviename.

Hmm, upon checking the manual, perhaps if Mac OS X lacks support for xv you should be using -vo quartz.

Mplayer has been under "development" for several years. It hasn't seen any major or even minor feature additions. /me boggles.

You've got to be joking. What about the latest ChangeLog [mplayerhq.hu]?

The user interface sucks, especially on OS X.

Actually, mplayer has some kind of bitmapped interface, which I always compile out. I hate all of those damned bitmapped interfaces, the pseudo-VCR things. It has a CLI interface which is exactly the same on OS X as on the other platforms.

Re:What happened to BitTorrent plan?

[Anonymous Coward]

From what I hear, it is being tested with around 1000 users right now. Its not just the archives though - its part of a larger project to have a 'Listen Again' (which was for radio) style system for TV.

Re:openness is hardly a concern to mplayer develop

[blibbler]

Sorry for the double post, but it needs to be formatted properly

Dirac + ogg in an mkv container could save your soul

I assume you mean Dirac + Vorbis in Matroska... Ogg is a container format like matroska... (what you said was akin to DivX + AVI in ASF :-p )
On a related topic, Matroska is an unfortunate format to use. It is very inefficient, both in CPU time, and in overhead. My computer can play a 640x480 divx in an avi, but can't play a 512x384 divx in matroska. Additionally, matroska has a higher file overhead than asf or quicktime.
Ogg is a good, basic format. It could be seen as a sturdy replacement for AVI. Matroska doesn't do anything beyond even the Quicktime format... and quicktime is vastly more efficient.

Anyway, moving on from that rant. I agree strongly with your other points. If you have a choice between two, largely equivalent codecs, one is unencumbered by laws and one is, why would you choose to use the one that is. Sure you might not get sued straight away, or even for a year or two, but why bother with that risk at all?
I am not particularly bothered with pirating certain commercial software, but if there is a free program that does everything I need, I will use that over a pirated of a commercial program. Why take the risk when there is another option?

The FAQ states...

[Kristoffer Lunden]

What about Ogg Theora?

We're glad that it looks like Theora will reach beta soon. We think you can't have too many free codecs, but that the Open Source community also needs to continue to develop codecs with increasingly better performance. We also think there could be a good deal in Theora that we could use in Dirac, and we'd like to work with developers who've been closely involved in Theora. We intend to pack the Dirac elementary stream into MXF, which has lots of useful features. That doesn't preclude it packing into Ogg as well, and it's probably a good idea to have a variety of packing formats. For this the elementary stream needs to be very well defined.

http://www.bbc.co.uk/rd/projects/dirac/documentati on/faq.htm#11 [bbc.co.uk]

Re:Is it in M*ENCODER*

[diracvideo]

Both a C interface to our encoder and direct show filters are on our road map. Monitor our site or check freshmeat to find out as soon as they arrive. Please bear in mind we are only in alpha.

Re:Patents

[diracvideo]

Dirac is licensed under the MPL (Mozilla) triple license. This means it is primariliy licensed under the MPL but anyone is free to re-license under either the GPL or LGPL licenses. We used the Mozilla license becuase it is well known and deals with patent issues. It means that any patents the BBC can license relating to the software are implicitly and irrevocably licensed for the software. So there are no royalties to pay for developing or using this software (either encoder or decoder). We chose this license to try to ensure that the codec remains royalty free. We figured if it was good enought for the Mozilla project it would probably be OK for us too. On the other hand we are aware that the MPL, on its own, is incompatible with the GPL and LGPL. We didn't want this and so have allowed re-licensing under the GPL. This means GPL software can freely integrate and develop Dirac. This is the approach suggested on the Gnu license page. You can also license Dirac under the LGPL. This means that even proprietary software can use it if they dynamically link with the libraries. The reason we have done this is so that the widest range of users can use Dirac royalty free.

Re:"licensed for free within the Dirac software"

[diracvideo]

Dirac is licensed under the MPL (Mozilla) triple license. This means it is primariliy licensed under the MPL but anyone is free to re-license under either the GPL or LGPL licenses. We used the Mozilla license becuase it is well known and deals with patent issues. It means that any patents the BBC can license relating to the software are implicitly and irrevocably licensed for the software. So there are no royalties to pay for developing or using this software (either encoder or decoder). We chose this license to try to ensure that the codec remains royalty free. We figured if it was good enought for the Mozilla project it would probably be OK for us too. On the other hand we are aware that the MPL, on its own, is incompatible with the GPL and LGPL. We didn't want this and so have allowed re-licensing under the GPL. This means GPL software can freely integrate and develop Dirac. This is the approach suggested on the Gnu license page. You can also license Dirac under the LGPL. This means that even proprietary software can use it if they dynamically link with the libraries. The reason we have done this is so that the widest range of users can use Dirac royalty free.

Re:The FAQ states...

[dolmen.fr]

For those who, like me, wanted to know more about MXF, it is the Material eXchange Format [snellwilcox.com] (and not "Multimedia Exchange Format" as Thomas Davies said in the interview).

when a licence is needed (& good for Dirac!)

[feepcreature]

No it isn't, it's illegal to watch the BBC without paying for it. If all you use your television for is console games or other channels, don't tune your television in to receive BBC channels and you will be fine.

The actual rule is that if you operate equipment that is capable of receiving broadcast TV signals you need a licence (UK spelling :-) ), or you have to license it. Even if you only watch commercial TV, or you never turn it on, or you only watch videos or play games (honest, guv).

There are minor exceptions for small, battery-only devices in caravans, and the like, and people have escaped where they could show that they could not receive broadcasts (like there was no aerial for their TV/Video, and neither were tuned in to TV channels), but it's not easy.

And the rules on multiple-occupancy are complex, and the way the BBC seeks to apply them seems to evolve. But the basic idea is that you need a licence if there are any tellies in a house.

It's enforced by automatic reminders, detector vans, £1000 fines (Grand if you have a licence, a grand if you don't), and TV sellers notifying the authorities when they sell a telly.

In my opinion it's great value, though. Even if it is being dumbed down, and the free internet-accessible archive seems to have stalled somewhere...

But developing a codec is sensible if it keeps the costs of streaming down - and especially if it's easier to implement on Open Source O/Ss. Hopefully some penny-driven accounting type at the Beeb won't feel the need to charge huge licence fees or impose Open-Source-incompatible licence terms.