"The three newer formats also outperformed MP3s when using the smaller bitrates typically used by websites to "stream" song samples to an individual's computer. But the advantage largely evaporates once files are encoded at bitrates that exceed 160 kbps, says Roberto Jose de Amorim, a computer science student in Curitiba, Brazil."

Read it here.

Unfortunately the article is really horrible They 'praise' the rigorious double-blind testing by "LAME developers" yet the article itself is full of unsubstantiated claims which clearly aren't based on solid testing.

Using LAME because iTunes AAC @ 128kbps is clearly artifacting? Hahah, sure.

I also like this one:



*Snort*

Oh, and how about getting the facts you are 'reporting' wrong?



I wonder what test that was that LAME won over Vorbis or AAC...

I'd surely like some of what that guy was smokin'

http://www.rjamorim.com/test/128extension/results.html - LAME lost against AAC



http://www.rjamorim.com/test/multiformat128/results.html - AAC was better, cannot say it won - but for sure it was not the other way around



So these are the facts backed by science... but it seems praising "same model as the Linux Operating System" sometimes has to win over them, eh?

what an odd article...

He's not the first person to have rather, well, odd, results in talking to Wired Magazine.

OMG! I'M FAMOUS!!!

Worship me mortals!




gah, not that outdated test again. I should take it offline. Everybody notices it before the newer test because it's higher up the page.


Besides, the article doesn't give enough information to judge, really. It only says he ripped his songs with iTunes and quality sucked, then ripped them with LAME and quality rocked. What if he somehow got 64kbps on iTunes, or even lower? And just the standard setting in lame - which is pretty, pretty good?


Edit:


Well, I didn't read the whole article yet, but he didn't misquote me there. I told him it is my opinion that past 160kbps, pretty much any codec - LAME, Vorbis, MPC, any AAC and even WMA Std - starts sounding "good enough" for the vast majority of mankind.

I think the writer was probably confused about the difference between using the LAME presets and using LAME at 128 kbps. For a lot of people, LAME is -preset standard (or now V 2 --vbr-new) and nothing else. And certainly that would beat 128 kbps AAC in most listening tests.

So, that first test was outdated... Did LAME rocked in any of your newer tests, while AAC sucked compared to it? Obvioulsy - no, it did not.



Wrong - he said:



That, then, reminds me on one similar test, organized by a... hmm.. one big company from the west coast How about the weight of the claims afterwards?

It would be complete irresponsible journalism to mention scientific methods, advanced math and computer science but blaming obviously more advanced technology for your own wrong or suboptimal encoding setting.

Jeez. That is too... lame

Puh-leeze! They were tied and I say that clearly.

"Lame seems to have improved a lot, getting tied to it's technological successor AAC."



Weightless as well, of course.



I agree that is lame. But I honestly have no hope of finding a journalist that knows as much about blind listening tests than ff123, Guruboolez, JohnV or me. We manage to live with what we have..

How true this is for average listeners that are not trained to hear codec-specific artifacts ... beyond the threshold of perceived transparency, it's basically a matter of +/- 20 kbps that separates MP3 from AoTuV, AAC or (insert any modern codec here) in my opinion.

Correct - they were tied. But none of these test could even remotely back-up the claim from the journalist that AAC @128 kbps sucked compared to LAME.

That is just ... wrong. Unless he used some broken AAC decoder.

BTW, they also mention boojum:


and timcupery:



Funny thing is, I talked with him for more than an hour, and that only amounted to a paragraph

I think I'll have to disagree here. There *are* design issues with MP3 that aren't fixable by a good encoder and with crancking up the bitrate.

Granted, there aren't going to be many clips that error out for an average listener at 320kbps, but for sure they exist, and perhaps the LAME team knows some.

AAC and Vorbis simply don't have the design "mistakes" MP3 has.

http://www.hydrogenaudio.org/forums/index....=37362&hl=wired

Considering that we almost took administrative action against him because we first thought he was a spammer (not because of that post)... heh.

I don't know if you are talking about me or the article writer. I was comparing 128kbps mpeg-4 AAC to -V 2 --vbr-new (then it was the --alt-preset standard) Lame 3.96.1 mp3. He did not put all of my quotes in. I told him that once I got up to 160kbps any format that I couldn't readily hear the difference between the formats. I always run my audio tests in my car as it has the best sound system. I hook my notebook up to my receiver and use ABX testing. I could easily pick out 128kbps mpeg-4 AAC and the original lossless source. However, once I got up to 160kbps it became hard but I could still pick out some tracks. Once I got up to 192kbps then I could not pick out any tracks what so ever (still in ABX testing). With general listening, even in my car, I though 128kbps mpeg-4 AAC was pretty damned good. That became an entirely different story once I turned my amps on (sorry, I don't have the ABX text results anymore, once I found my format and preset then I deleted them).

So, that was just my ears with my music and my equipment. Look at ALL the tests that use Lame -V 2 --vbr-new as the higher ancor, it pretty much always outperforms the 128kbps bitrate formats.

Edit: I forgot to mention the albums that I used for testing. Korn - Issues, Lamb Of God - New American Gospel, Pantera - Far Beyond Driven, Tool - Lateralus, and Judas Priest - Metalogy (Disc 1).

kraftwerk...

Which is exactly the reason I think the article is not good.

- Some very important facts were cut out
- Codecs were compared on different bit-rate scales
- Results of one person, that probably would not correlate with the average, were translated into persuasive story.

Nice

While most people with PC's don't think that Audio CD's contain WAV files, I find that nearly all Mac users think that Audio CD's contain AIFF files. Why? Because Macs display AIFFs when you stick an audio CD in them, you can copy the files to the computer as if they were AIFF files, and you can copy AIFF's to CDRs and make Audio CD's in that way. So as far as a non-knowledgable Mac user can tell, Audio CD's are just CD's with only AIFFs on them.

Congrats to the dudes that got mentioned.

The essential thrust of the story seems accurate enough though:

1. Music available from commercial online sources is generally of bad quality.
2. Mp3 is still by far the most widely supported codec.
3. The Lame team's work has made it possible to encode music transparently for most people for most music at very reasonable bitrates, diminishing the appeal of alternatives substantially.

HA readers might care about the design flaws of mp3 but the typical listener won't as long as their music sounds good.

I too thought that the essentials were understood. I thought the article also did a good job in praising the Lame developers for the good jobs that they have done with the Lame mp3 encoder. Sure, the mp3 format has design flaws by nature but as long as the music sounds good then it is worth it especially to get the universal hardware playback options that mp3 has. Good job to the Lame developers and congrats on the praise in the article.

The gist I got from the article was that MP3 is still the most supported, and sounds pretty flawless at high bitrates. At lower bitrates other codecs have the upper hand.

That's correct, no? The author never makes any specific claims about CODECs at specific bitrates. In fact I'm impressed that he mentioned double-blind testing without mentioning all the irrational detractors. The only glaring mistake was saying that CDs contain .wav files.

PC people don't think audio CDs contain .wav files; they think they contain .cda files.

Master, Master
Give me Vorbis, Lame, and faster
Master, Master
Encoding all my files

They mention iTunes 128kbps AAC specifically. If you think that is bad quality, please take Sebiastian Mares' test and come back afterwards, thankyouverymuch.

No, as was already pointed out.



You're right: he makes even worse general ones.

I've read the entire article, and IMHO, it is among the better written "mainstream"-articles about lossy digital audio on the net.

Sure, not every detail is right, and the reader gets the impression, that some advantages of LAME (like the DBTs) are exclusive to it, and less applied to the other codecs. And a few minor details of course. However, overally, it gets more things right than it does get stuff wrong - and thats by far not usual in the mainstream press. And the basic premise of the article - from a consumer POV - is absolutely right: For most people, the advantages of other codecs are simply too minor to outweight the overwhelming support and compatiblity of LAME-encoded MP3s... with the only exception being narrow-band (i.e. webradio) scenarios.

YMMV - especially if you're a developer of an encoder other than lame :)

- Lyx

I've got quite a few 128kbit iTunes AAC files because I transcode to that format for my iPod. I don't think it sounds "bad", but it's certainly not transparent and there's no way I'd buy music at that bitrate.

If you guys think this article is that off base then you should write Wired and ask for a correction. I still think that the essential points are basically correct, particularly from the point of view of the typical Wired reader.

Couldn't agree more.
Well, it's not like I wouldn't defend a product that I make myself.
Especially one that I really take some pride on.

Irrelevant i know, but does anyone know offhand what endianness is CDDA? I think .aiff is big endian. OS X has flitted and changed between how it displays the contents of audio CD's. In 10.2 i'm sure they were .cdda files which were actually different to the .aiff it presents them as now. (you needed SoX or something QuickTime based to read them) Interesting, nonetheless.

Good point, and I would certainly not call "audiophile-ready" a format that clearly cannot encode some clips without audible distortion, regardless of the bit rate or the codec settings you use. And that format is not AAC in this story.

Let's digest some things from this article:



Well - AAC does, in fact, outperform MP3 in "bit to bit" - on the same bit-rate AAC is almost always better than MP3.

Now, the fact that someone wants to compare 128 kbps CBR AAC with MP3 of much higher average bit-rate is a completely different story - I don't think anyone claimed that AAC at 128 kbps CBR could be better than, say MP3 @180 kbps VBR for a typical sample set. This is just wrong comparison, and I am quite sure the person who was quoted knows that very well. It is just the journalist who translated that into a misleading paragraph.

Now, hiding that obvious flaw, author goes even more with "scientific stuff applied in LAME" such as DBT, Computer Science and other buzzwords such as allmighty "model as the Linux Operating Systems" - like other codecs were not developed by using the same scientific means... Sheesh.

Finally - the "Ordinary Wired Reader" would get an impression how LAME is actually better than AAC - and how it is developed with uber-cool scientific methods

This is the core of the "mis-leadingness" of the article - I could only imagine a huge flamewar here if someone compared 128 kbps CBR LAME with, say, 180 Kbps WMA VBR and started talking about "modern scientific approach applied to blahblah powerful algorithms used in WMA" - leave alone some well-known magazine. It would be one hell of the flame

And... for the case of defending "free as in freedom" codec, it seems that people are willing to get over all that

I wasn't planning to mention it, but i guess the above quote makes it almost inevitable.

Yes, you are right that this comparision is unfair.... however, it seems that the marketing department of nero isn't that much interested in "even" comparisions...

From http://www.nero.com/nerodigital/eng/Nero_D...highlights.html :
- "In fact, CD quality stereo at 48 kb/s"
- "Transparent quality at 128 kb/s. MPEG-4 AAC provides transparent audio quality at 128 kb/s according to all professional and formal listening tests conducted so far. At 96 kb/s MPEG-4, AAC offers quality of 128 kb/s MP3."
- "MP3 quality with 50% of the space. With High Efficiency AAC in MPEG audio, you will get same quality from an MP3 with 50% of the space"

Funny thing is that unless one fills in the blanks, some of the above claims by nero even contradict each other - if one takes them as generalized as they imply to be.

Now, i know that two wrongs dont make a right - but i just thought a friendly reminder may be in order.

- Lyx

You are correct. I did bit for bit comparisons of mp3 and AAC. After 160kbps VBR, I could only hear the difference, when doing my blind ABX testing, when turning the amps on in my car. Only then did the mpeg-4 AAC format get slightly better scores than -V 4 --vbr-new Lame mp3. I then tried -V 2 --vbr-new (at the time it was --alt-preset standard) and 192kbps mpeg-4 AAC. I could not readily hear the differences between the mpeg-4 AAC files and the Lame mp3 files. Keep in mind that I am comparing the ABX results (when comparing the lossy file to the original wav) of each format. I am not doing a ABX test where A is the mp3 and B is the mpeg-4 AAC. After doing my testing, I then found that 192kbps mpeg-4 AAC and --alt-preset standard Lame mp3 offered transparency for me for all samples that I tested (I tested about 10 tracks). It was then that I realized that I might as well use --alt-preset standard as the file sizes were either slightly larger or slightly smaller than a 192kbps CBR mpeg-4 AAC. Additionally, I found the universal hardware media playback support that mp3 has to be astounding when compared to the hardware playback support of mpeg-4 AAC, at the time.

I think we are missing the timeline - different profiles of AAC were used for different claims, so I see no contradiction unless we forget that important fact while "filling the blanks"

Also... I think we already over-discussed Nero website claims:



Actually transparent might not be the perfect word, but "indistinguishable from original" according to the ITU terms, and proved by MPEG listening tests.

So the wording might be a mistake according to somebody - but for sure there is not a word mentioning statistical transparency which is something quite harder for a codec to achieve (what MP3 cannot achieve even at 320 kbps, for instance)

96 kbs AAC vs. 128 kbps MP3 was also a fact proved by other formal listening test - I have seen no big listening tests disputing this claim anyway - if you wish to set up some tests, it would be very nice

Right, and thats exactly what i tried to point out: You complain about the wired article being misleading because important info is missing. Nero on its website is doing exactly the same - certain info was intentionally left out, so that the reader gets a wrong and misleading impression of the codecs performance. If you want to be treated fairly, tread others fairly.

Another point which can be derived from the above is something which i mentioned earlier, and which kornchild explained in more details again: Why does Nero feel the need to paint an exaggerated picture of their encoder's performance? Would they need to do that if their encoder would indeed be so much significantly better that it makes up for the far lower software/hardware support/compatibility? Well, thats exactly the problem of all those "modern" lossy codecs: Besides of narrowband, their advantages are just not significant enough for most people to make up for the worse support.

Would a normal user give up the "play-everywhere"-support of mp3, for 10% lower filesize? No. 20%? No. 30%? Maybe. 40%? Probably, but it just isn't so.
MP3 is "good enough" for most people and the newer codecs advantages simply cannot deliver a significant enough gain to justify switching to them - so the company's behind them resort to exaggerations instead to "justify" their use.

But when a user or magazine then take's those exaggerated suggestions as "true" and judges them by those criteria - then they complain about "unfair comparisions". And here, we come full-circle...

- Lyx

Perhaps because of a small difference - the purpose of the journal/newspaper/magazine is to inform the public about all products/technologies while the purpose of the company PR, regardless of the company, is to promote company-specific products/technologies.

For me, it is a very huge difference, huge enough to justify complaining about obvious flaws in the magazine article.

Thanks for the fair and honest reply.

- Lyx

Wow...CD quality used to be 64kbps with wma, but aac can deliver that in 48kbps, cannot wait for next year when CD quality will be 32 kbps... [sarcasm off] seriously it is propaganda to represent something that is just 3% the size of the original and call it CD quality, standards in the audio industry are slipping.



Ok, not disputing that one but "MP3 quality with 50% of the space", so for your above mentioned test a 96Kbps AAC would have to equal a 190Kbps mp3 file, is there such a test? I know the answer (32 kbps aac probably equaled a 64kbps mp3 test) but 64kbps is not what mp3 is about and that claim is slimy, it is like claiming a tank as fast as a Ferrari, perhaps it is if you put them both in pool of custard but it not an absolute truth.

You're going to love this, but it depends on the CD Drive you are using. Some drives return little endian data, some return big endian data. The ripper has to account for that by asking the drive what byte order it uses. Most CD drives return little endian nowadays, I think.

You are correct that AIFF is always big endian.

>Some drives return little endian data, some return big endian data.

Only drives that were 1xspeed from 10 years ago (when there were proprietry raw read commands), all other drives unaffectted.

I agree. Other than the obvious pro-MP3 propaganda that was bound to displease some people around here (and which was greatly my fault, as I went to great lenghts to convince him that MP3 is the way to go for anyone but people in specific situations - E.G, iPod owners), I think his article is mostly correct.

Also, I talked to him on Skype yesterday night, and told him to correct a little part. "The .wav file format, found on CDs issued by record labels" was replaced with "The PCM format, found on CDs issued by record labels". That was the only real error, IMO.

A small, yet big difference.
From what I read here, it seems that when you go promoting something, it justifies almost everything.

But, forgive me if I misunderstand something...

I don't really think I ever stated that - but if we speak in general case, the purpose of marketing is to promote the product/service - and bias is not something uncommon in that science.

It is absolutely not true that marketing implies "almost everything" - but usually marketing focues on communicating the features/advantages of the products compared to competition.

This is why people tend to trust journals and magazines for complete overviews and comparisons of the technologies/products/services - and this is exactly why mistakes in journalism are much more problematic than structure of some marketing campaign.

But this is just by 0.02$ ...

I've thought about it before. It would be very interesting to test all these claims of "CD quality" or "equal to 128 kbps mp3" or whatever is hot at the moment... Four codecs, which have all been marketed by bold claims, that I would include:
Nero HE-AAC v2 @48 kbps
WMA Standard @64 kbps
(iTunes) AAC-LC @96 kbps
Coding Tech MP3 Pro @64 kbps
vs the high anchor
(LAME or FhG) MP3 @128 kbps

All three have been marketed as "as good as" or better than MP3 @128, but I have my doubts... What happened with the test Garf was planning for? It died?

Edit: added mp3 pro

It faced credibility concerns (remember that PM swap we had?)

But now that I finally found someone to replace me, nothing keeps Sebastian to give such test a try. It would indeed be extremely interesting.

Edit: I would also add MP3pro at 64kbps to that list.

Replace you???? Not possible

Edit: Congrats on being mentioned in my favorite mag!

I remember. But I thought he would have more perseverance than to give up for that reason...

Will he replace you as the sharp tongue also?

I forgot mp3 pro, yes... And I've thought about Vorbis and Real but can't remember any such bold claims from them. It says Vorbis is better than mp3 at a lower bitrate but not by how much. Real I simply haven't taken much notice of.

Hehe, thanks

I meant that as "Replace me as listening test conductor"



Doesn't seem to be the case. And ultimately, that's why the pre-test thread has gotten so awfully awry :B

I would just have told some blockheads to STFU and go conduct their own tests.



In the RealJukebox days, Real claimed 96kbps RealAudio = 128kbps MP3. Dunno if they later dropped that claim. Then again, they used early Xing for MP3 encoding in RJB, so they might have been correct after all - if you considered their app alone.


Xiph indeed never made official claims of "same quality at x% smaller bitrate". Kudos to them for not having to drop to such misguiding marketing plots.

The fact that most of the times marketing claims suck is no excuse for a supposedly serious magazine to lead to inaccurate and totally wrong impressions, such as mp3 being better than AAC. A whole different thing would have been saying that that in practice mp3 can be good enough for most people. That would not have been very objectionable.