Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: AAC versus LAME: Recent Tests from 2009 (Read 12601 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

AAC versus LAME: Recent Tests from 2009

Just having a look through using the search function - the last topic I found was back in 2005. Has there been a recent comparison between AAC and Lame - AAC using Apple iTunes/Quicktime versus Lame 3.98.2?

I've resided myself to using MP3 instead of AAC for portability between players (media players and portable players) - wondering whether there is a significant difference once one stars using a high setting like -V 0 versus cranking it up to the 256Kbps 'iTunes Plus' setting iTune defaults to now.

AAC versus LAME: Recent Tests from 2009

Reply #1 –
The general formula is for a share of x>90 % there is no audible difference. For a share of y<100-x % (problematic samples) AAC (Nero, Quicktime) beats MP3 (Lame, Fraunhofer). There is no known sample where MP3 would beat AAC. For the smallest share z<y (killer samples) both fail.

This formula hasn't changed much recently.

AAC versus LAME: Recent Tests from 2009

Reply #2 –
Quote from: rpp3po on
The general formula is for an amount of x>90 percent there is no difference. For an amount of 100-x percent (problem samples) AAC (Nero, Quicktime) beats MP3 (Lame, Fraunhofer). There is no known sample where MP3 would beat AAC.

This formula hasn't changed much recently.


I'd assume, therefore, once you start getting up to around around the bit rates I am using, there is very little difference?

I've been listening to a variety of music compressed using lame -V 0 and found that there is either no difference or lame produced superior audio quality.

AAC versus LAME: Recent Tests from 2009

Reply #3 –
Shares y and z contain samples that fail regardless of bit rate (even at V0 or CBR 320). This is more often the case for MP3 than AAC, but in comparison to the vast amount of music out there, both shares are very small.

AAC versus LAME: Recent Tests from 2009

Reply #4 –
rpp3po, are you certain about the AAC and MP3 bit? Could you cite some sources? Both are lossy formats that require tuning, and LAME has been very extensively tuned. As different encoders use different tunings, I would expect them to fail in different places. Admittedly, as MP3 cannot provide bitrates beyond 320kbps, AAC has the possibility to be transparent when MP3 cannot. I'm not sure it necessarily follows that AAC problem samples will be MP3 problem samples.

AAC versus LAME: Recent Tests from 2009

Reply #5 –
I successfully ABX'ed AAC (using Nero Digital 1.3.3.0) at -v 0.4 (=132kbit AAC LC) with Enigma: The Eyes of Truth. Not sure if this is normal or just a rare occurrence, however I would expect an AAC at this bitrate to be tough - Isn't it the bitrate iTunes always used to sell theigh music on?

Just as a reference I was not able to ABX the same track with LAME -v5 (154 kbit).
Can't wait for a HD-AAC encoder :P

AAC versus LAME: Recent Tests from 2009

Reply #6 –
Then again, Lame had a 22kbps higher bitrate than Nero at -q0.4.  It is understandable that Lame produced better results when using that "high" of a bitrate compared to Nero AAC.  iTunes used to sell their music at either 128kbps CBR or 128kbps VBR constrained, there was no clear indication which songs were which.  They then started to add 256kbps CBR and 256kbps VBR constrained files and I think that they only sell 256kbps VBR constrained songs now.

AAC versus LAME: Recent Tests from 2009

Reply #7 –
Quote from: kornchild2002 on
It is understandable that Lame produced better results when using that "high" of a bitrate compared to Nero AAC.

I was just under the impression that AAC were far superior to MP3, not on par
Can't wait for a HD-AAC encoder :P

AAC versus LAME: Recent Tests from 2009

Reply #8 –
Maybe at very low bitrates, but not "standard" DAP/home listening bitrates.
I'm on a horse.

AAC versus LAME: Recent Tests from 2009

Reply #9 –
On killer samples Nero AAC is usually far superior to Lame MP3, it's just that people using Lame MP3 like to fake their mind thinking that MP3 is just as good as AAC. It is not true.

They only extrapolate on the fact that on non-killer samples MP3 sound as good as AAC: this is a wrong way of thinking. On non-killer samples, (providing you use a modern codec & a bitrate superior to 192Kbps VBR) 90% of music sound the same no matter the codec you use. (unless you use AC3, VQF or Blade ... or 96Kbps)

You shouldn't listen to people using lossy when searching advice, unless they provide serious ABXing log, they are biased. A few years ago almost half of the users from HA would have told you how great Musepack was (No comment). Even with serious ABXing, audio quality is not everything when you judge a codec. No matter how good it sound a codec without developper is a dead codec (even with an open source licence). There is nothing wrong with Lame MP3, but if you use it, you do it due to its superior compatibility, not due to its superior quality.

I use 100% lossless (except lossyflac on DAP but I don't keep the encoded file long), I have made ABXing for myself, Nero AAC is the lossy codec that provide the overall best quality at 192Kbps VBR, & despite that actually, I don't use Nero AAC myself.

I don't particulary like Nero, Nero codecs will never be open sourced, but when it come to quality Nero AAC has the lead. When you know it for sure because you tested for yourself, you have to admit it.

Before I tested I thought Vorbis was roughly equivalent to Nero AAC, I was faking my own mind too. Like Lame MP3, Vorbis is far from being as good as Nero AAC on killer samples. Indeed on non-killer samples Lame MP3, Nero AAC & Vorbis all sound the same. I mean you cannot be more transparent than tranparent, a transparent Lame MP3 sample is just as transparent as a transparent Nero AAC sample, does it mean that Lame MP3 is as good as Nero AAC ? NO.

In my old listening test I made Vorbis look very bad as I selected killer samples that were targetting it badly, I could have made Lame MP3 look just as bad by adding samples where Lame MP3 fails badly.
But the fact is that I couldn't have hurted Nero AAC as badly even if I had wanted to do so, Nero AAC is just more resilient & it still improves.

That said Nero AAC, (like all lossy codec including lossywav) is far from perfect.
When you get it, you use flac

A lot of people seems to extrapolate Nero AAC encoder result to ITunes AAC encoder, this is wrong too. I tested Nero AAC a lot & I cannot say anything about Itunes AAC encoder, I just didn't test them.

As for at which bitrates Nero AAC beats other codecs, at 192Kbps VBR nero beats all other codecs but the artefacts of other codecs are usually light at 192Kbps VBR, so I would say that starting at 128Kbps VBR & below, Nero AAC beat all other codecs with a margin that is so high that using any other codec is sub-optimal if you only consider quality (you can still use vorbis for its license or mp3 for its compatibility).

So Nero AAC is the best starting at mid-low bitrates, & not only starting at really low bitrates like it is usually spreaded. That said there is no lossy codec fully transparent at 128Kbps VBR even Nero AAC.

Some people lie to themselves thinking that Lame MP3 have been "so carefully tuned over the years" that it would rivalize forever with Nero AAC, this is just wrong. Nero AAC beats Lame MP3 & it beats it right now, just as badly as x264 beats xvid, it's just that in the audio world the difference is not as obvious as in the video codec world.

AAC versus LAME: Recent Tests from 2009

Reply #10 –
A couple of points from your verbose post:

Quote from: sauvage78 on
As for at which bitrates Nero AAC beats other codecs, at 192Kbps VBR nero beats all other codecs but the artefacts of other codecs are usually light at 192Kbps VBR, so I would say that starting at 128Kbps VBR & below, Nero AAC beat all other codecs with a margin that is so high that using any other codec is sub-optimal if you only consider quality (you can still use vorbis for its license or mp3 for its compatibility).
I realise that the test was a while ago (I'm sure both codecs have been tweaked since then), but this is not what the public listening test found.  You may have golden ears, but most do not.

Quote from: sauvage78 on
Some people lie to themselves thinking that Lame MP3 have been "so carefully tuned over the years" that it would rivalize forever with Nero AAC, this is just wrong. Nero AAC beats Lame MP3 & it beats it right now, just as badly as x264 beats xvid, it's just that in the audio world the difference is not as obvious as in the video codec world.
So, which is it?  NB: I do not disagree that AAC is a superior codec, but I do disagree that it can achieve "far superior" results to MP3 at mid-high bitrates (and I suppose that depends on your definition of "mid").
I'm on a horse.

AAC versus LAME: Recent Tests from 2009

Reply #11 –
Nero AAC achieves "far superior" results on killer samples, not on usual music.
(In my mouth, 96Kbps is low bitrate, 192Kbps is high bitrate, 128Kbps is mid-low, 160Kbps is mid-high.)

I don't consider myself to have golden ears, because I consider that if most people would take the time to listen carefully they would most likely find the same result as me.

There is not the golden eared on one side & the non-golden eared on the other side.
There is the people who take the time to listen carefully & the people who don't.
Having good ears is less important in ABXing than patience. ABXing is long & boring even with decent earing.

The public listening tests are ... the public listening test, I will not say anything bad about it except that it shows an average. Personnaly, I don't judge the quality of audio codecs on the public listening test anymore, not because it is bad but because among the people who test there are advanced user who know what they listen to & complete newbie who don't know what an artefact sounds like. Furthermore I couldn't identify all problem samples as real problems samples. (Not that they aren't problem samples for others, just that I prefer focusing on sample that I can identify as problem samples for me)

Personnaly, I rely much more on personnal listening test of other individual beings, I ask them what do they hear & when, then I try to see if I can find the same result. Only then I trust their result. If I hear nothing I don't draw any conclusion. It is a much more time greedy processus than just trusting the public listening test. But I find it more reliable.

The sad thing about it is that it pushes you to use lossless, because the more you test lossy, the more you realize what you lose. It get even worst when you realize the time you wasted doing ABXing.
Lossy is not only a waste of audio quality, it is a waste of ABXing time.

Edit: no h to ears !

AAC versus LAME: Recent Tests from 2009

Reply #12 –
Problem is people expect too much from audio codecs designed to reduce data rate and sound good (not perfect)

To achieve 7:1 or more compression these codecs must hide a lot of error signal and the do it good most of the time. Lossless can achieve around 2:1. In between you can get perfection all the time but must settle for around 3:1 with a hybrid encoder. This works even without any psy modeling whatsoever at any sample rate - at least to my ears. for high rez audio there should be significant savings.

AAC versus LAME: Recent Tests from 2009

Reply #13 –
The ABX time waste is a real issue. I attribute this to the digital era we live in. Too many choices & complexity  - none are satisfactory. Remember tape, CD's that played in everything gapless, radios with two simple knobs, TV's , etc - Life was easier and i think many people were happier with having less choice.

Back to the subject, I think the move to lossless (in HA) is the right direction - as a single unified library whenever possible. I long for a high quality or lossless simple solution that just works. I wish mp3 would have just become a legacy thing and a HQ / lossless codec (a standard) would take over with proper gapless specs across all players.

MPC was mentioned as dead. Truth is I would have been very happy even today with MPC -standard if it would have had a small audiophile market and all devices that supported it would play gapless outside rockbox.

AAC versus LAME: Recent Tests from 2009

Reply #14 –
Quote from: Canar on
rpp3po, are you certain about the AAC and MP3 bit? Could you cite some sources? Both are lossy formats that require tuning, and LAME has been very extensively tuned. As different encoders use different tunings, I would expect them to fail in different places. Admittedly, as MP3 cannot provide bitrates beyond 320kbps, AAC has the possibility to be transparent when MP3 cannot. I'm not sure it necessarily follows that AAC problem samples will be MP3 problem samples.


Last time I asked him he did not have any sources.