Conclusion

The qualification isn't about finding the best codec, it's about finding those that can stand up in the main round, against the best in the field. Thus, I cannot really tell you which codec performed best in the qualification, I did not put as much effort into it to be confident in making such a statement.

However, I'm confident in saying that if I had done another pass and only squaring off Elecard and XviD, that question would be answered. In all other cases, I felt XviD had an edge.

In light of the new two tiered main round, I'll compare the qualificants in their respective groups.

In the AVC group, Elecard clearly was the best performer. It delivers reasonable speed, reasonable handling (obviously the commandline encoder would be a bonus for people who like to get down and dirty), and caused me no problems in the comparison.

Unfortunately, the same cannot be said for the second AVC codec. Even though VSS' offering now supports the high profile, the rate control is still lacking, making it unsuitable for many purposes. I said it last year, it's something that can be worked out, but I wished a little more effort was put into this issue. High profile is nice, but if your ratecontrol has such big problems.. You're not going to be in the main round.

QuickTime leaves a mixed impression. It is basically a baseline profile codec with an added b-frame, but just that, and it shows. It is also excessively slow, and usability is frankly a nightmare. If there's one thing you can take for granted, it's that you will find very few people willing to save entire movies in raw YUV format just because an encoder cannot handle VfW or DirectShow input. If you want a simple profile encoder, there are better options. And while I'm writing about QuickTime, I also need to mention its playback capabilities, or lack thereof. Just as a video encoder, its playback capabilities are severely limited. Yes, it can do Main Profile, but only with placing 10 additional restrictions that are nowhere in the specs upon you. So before you start blaming your favorite AVC codec for QuickTime's inability to properly play it, change the player. Quality is rather low level and usability is way below acceptable, so this codec will not make it to the main round. And as a sidenote, Apple chose not to reply to my request for help with the comparison, so if you have any objections to settings, please send them to Apple, maybe they will bother to reply to me next time.

Then we have the ASP group:

XviD will make it very hard once again for the competition in the main round.

LMP4 can compete closely in still scenes, but the extended action scene showed a clear difference in favor of XviD. Still, I feel a definite improvement over the last time I tested the codec, so it will advance to the main round.

NeroDigital ASP has been improved rather significantly over the last time it competed (RDO does have its perks), and the codec is really lightning fast, but it's going to be really though in the main round.

Finally we have the non MPEG-4 group:

If it weren't for the discoloration issues, Snow is the best proof that wavelet based codecs can work, and work properly. So I don't think I'm the only one hoping Michael Niedermayer will resume work on the codec, so that it will make for an interesting contender in future comparisons. But for now, I cannot allow this codec to pass to the main round.

As far as Theora goes, quality isn't up to the level and neither is rate control. I put this codec through the quality evaluation purely out of curiosity, it would've not been allowed in the main round because of the severe oversize anyway. Also, since the promised DVD backup optimized ffmpeg2theora build never showed up, I ended up using what's available per default and if you have any quarrel with that, direct your objections to the people in #theora on IRC.

Finally, Dirac didn't make it very far I'm afraid. At this point, the codec needs a lot of work to become a competitive solution for DVD backups / archival.

 

So there you have it, the Elecard AVC encoder, along with LMP4 and NeroDigital ASP go to the main round. And as far as the competition goes, I hope this qualification serves as a reference point where work needs to be done to get competitive.

Best quality per fps

Even with a dual core machine, you can see by the FPS numbers on page one that while there are fast codecs, there are those that could use a good kick in the behind to get moving.

It's really hard to say this year. With almost 50% over XviD, NeroDigital ASP gets the crown, though I must say these FPS you trade when switching to XviD seem justified.

Future outlook

Well, the have a look at the main round results.

Now if you have any questions, please have a look at the FAQ.

This document was last updated on December 23, 2005