http://forum.doom9.org/showthread.php?p=1128872#post1128872

Here developer claims that WMA10pro 48 kbit/s outperforms AC-3 CBR 160 Stereo.

Have anybody information about it?

Just to clarify, I'm not a developer of the WMA Pro codec. That technology was developed before I was a member of Microsoft's codec team.

I know what Ben had in mind with this:

*AC3 is not a terribly efficient scheme when encoding only 2 channels, that's for sure.
*He thought about a 2 passes uncapped VBR resulting in overall bitrate of 48kbps in the context of a full movie soundtrack (and that includes a lot of easy parts).

To me this, even if unproven, might be plausible.

But Ben, the way you worded it is awfull. This sounds like an overly broad statement, while you had a specific use case in mind. Be carefull: wording it this way might generate some anti-Microsoft feelings...

Sorry if I was unclear; I thought I had clarified it accurately. Your analysis is spot on. I'm specifically making a comparison between:

• WMA 10 Pro 48 Kbps VBR 48 KHz stereo versus AC-3 160 Kbps CBR 48 KHz stereo
• With typical movie soundtrack content, with typical variability in content

1. Who put this kind of restriction like WMA10pro VBR 2 pass vs CBR AC3?
If you want a fair test then go both CBR or both VBR. AC3 has VBR mode too.

2. Actually "typical" has a very wide range. There can be a lot of ochestra, classic music, some background sounds etc.

How WMA10pro SBR like codec can achieve a good quality when all frequencies higher than 11 khz are synthesized by SBR part when AC3 160 preserves frequencies up to 16-17 khz?

benwaggoner, your statements like "internally test" and "my own anecdotal experience" are poorly informative and lead to disinformation.

I want to see public test.

If you go back to the original thread, it wasn't a discussion about WMAPro10 v. AC-3, but about downloads v. optical disc, and why you can get away with fewer bits for the same quality with a download versus optical. One of the reasons is that all the lossy audio codecs on optical are CBR.

Certainly AC-3 could perform better were in done in VBR, and WMA Pro 10 would do worse if it was used in an optical format that required it to be CBR. Again, my quote that we're discussing was about download versus optical, with codec options just one of the axes in that comparison.


Depends on how well the frequencies are synthesized, and what other artifacts AC-3 @ 160 Kbps gives. Also, FWIW I was assuming we were talking 48 KHz, although I doubt it'd matter much either way.


Sure. This is for the publicly available codec. You're more than welcome to run your own test; I'd be happy to hear what your results are.

I can make some files as well, but no one is claiming some secret technology here.

And the big one is that MPEG2 is terrible compared to more modern codecs.
Anyway, a modern VBR codec at 48 kbps VBR (using a setting that probably averages more like 80 kbps on actual music) beating AC3 at 160 kbps CBR isn't that surprising to me.

EDIT: Whoops, didn't realize that this thread hadn't recieved any posts in almost a year. Sorry for bringing this thread back from the dead.

Wow, blast from the past. I was searching for a different thread and found my past self .

I actually ran the head-to-head with Elephant's Dream ages ago, and it looks like I forgot to post the final files. I wonder if I still have them somewhere...

benwaggoner

I understand that it's part of your job to promote Microsoft codecs as WMV9/VC1 and WMA. But to do it you should give something to people: higher compatibility or higher quality. Your codecs haven't anything of that.

WMA isn't better than AAC or Vorbis at any bitrates http://www.listening-tests.info/
and there is no compatibility as for MP3. WMA is utterly unpopular http://www.hydrogenaudio.org/forums/index....showtopic=68338

The same situation around WMV9/VC-1. It's comparable only to old generation MPEG4 ASP like Xvid.
You have admitted VC-1 was inferior to state of art H.264/H.264+/++ videocodecs.

So what news can you actually bring to video and audio compression areas?

One thing you might want to remember is that the population of Hydrogen Audio is not similar to the general populace and that even if it was the poll would suffer from self-selection bias.