LAME MP3 vs iTunes AAC, Quality question Options

[NeoMoose]

Okay, back a month or so ago when Roberto did his listening test @ 128 kbps for all the formats LAME and iTunes essentially tied. Can someone please give me a quick explanation how LAME, based on the MPEG-3 standard somehow tied with iTunes which is based on the MPEG-4 standard?

Am I missing something here? I would think the newer MPEG standard would crush the bejeezus out of its predecessor. What's the deal?

[Tec9SD]

Can someone please give me a quick explanation how LAME, based on the MPEG-3 standard...

MPEG-1 Layer 3 standard.
The MPEG-3 standard-in-progress was disolved.

Now, I part for the experts to answer.

Edit:
Also, IIRC, because of the way ISO/IEC standardization process is, newer standards can be designed with different goals in mind, not necessarily as a replacement for the previous standard.

FYI: MPEG-4 is ISO 14496, I believe in case you'd like to research it.

MPEG-4 - Wikipedia while you wait for someone better to answer your question.

Edit 2: It's easier to find the answer when you know the right question to ask, that's why I corrected you. That's all.

This post has been edited by Tec9SD: Jun 18 2004, 08:12

[NeoMoose]

Well, at you know what I mean.

[Teqnilogik]

Well, LAME is a very fine tuned MP3 encoder that is still actively developed. So I'm not surprised that it has kept up with iTunes's AAC encoder at 128 kbps. I personally have encoded files with iTunes's AAC encoder and LAME at 128 kbps and I will say that LAME definitely sounds very close to AAC at 128 kbps but AAC still sounded a little better to me. I noticed more 'ringing' artifacts in the 128kbps MP3 than I did with the AAC so the AAC sounded better to me. Don't take that as a scientific result as I did not do any ABXing, those are just my personal observations.

In the future, I expect to see AAC encoders pull away more from MP3 encoders and perform significantly better. At the moment they haven't been developed long enough to be lightyears away from LAME MP3 quality-wise.

This post has been edited by Teqnilogik: Jun 18 2004, 08:16

[Gabriel]

MPEG natural audio standards (like mp3 or aac) are only defining the bitstream and the decoding, but not the encoding.
It means that different encoders can encode in a different way, with different efficiencies.

It is like books authors and language. MP3 is using an old language, and some encoders are making good use of this language, as if an author was able to produce a very nice book.
Now, you have a new language (aac) which allows you to express more subtle variations of concepts and wording. You know it is more powerfull, but book authors do not know it enough yet to be able to produce better books than with the old mp3 language.
However, you know that when they will learn more and practice more the new language, they will be able to produce better results than with the previsou one.

This is just a simple transposition of things, but things are similar with mp3 and aac. AAC is more prowerfull and more efficient in theory. But the results are quite dependant of how the encoder is using the standard.

BTW, I am quite sure that AAC at 128kbps could be way better than mp3. It just needs a few more time.

[moi]

BTW, I am quite sure that AAC at 128kbps could be way better than mp3. It just needs a few more time.

Interesting observation, coming from the LAME creator. (I might not be correct about that. I'm new at this. Are you the creator of LAME, Gabriel?)

Would the same statement apply to WMA, OGG, and MP3Pro, as well as AAC? Aren't they also variants of MP4?

Are you saying that you think MP4 based compression will still improve much more, that even though you are working on improving MP3 compression with new versions of LAME, that it cannot go as far in improving as MP4 based compression can?

If so, are you beginning to work with MP4 compression as well? If MP3 compression will become obsolete as Mp4 compression undergoes more improvement, will you continue to work on improving MP3, or start working on MP4? Have you started that? With which format?

MP3 is still by far the most compatible format, playable on all devices and systems. WMA is the second most compatible format, playable on most devices other than Ipods. OGG and AAC are far behind, although AAC exists on IPods, the most popular digital audio player. What do you predict about the future of these standards? Will AAC, or another MP4 format, eventually replace MP3 in compatibility, as the new standard? (I know Apple will probably not allow other hardware companies to use its version of AAC, but I understand there are other versions of AAC.)

[rjamorim]

Are you the creator of LAME, Gabriel?

The creator of Lame is Mike Cheng, and for a good part of it's life the project was maintained by Mark Taylor.

Gabriel Bouvigne is probably the most active developer working on Lame currently, next to Takehiro Tominaga.

Would the same statement apply to WMA, OGG, and MP3Pro, as well as AAC? Aren't they also variants of MP4?

Not at all. From the formats you mentioned, only AAC is related to MPEG-4.

[Teqnilogik]

(I know Apple will probably not allow other hardware companies to use its version of AAC, but I understand there are other versions of AAC.)

The AAC files that iTunes rips from CDs is not proprietary to Apple. They'll play in any standard MPEG-4 AAC player. Apple sells DRM-protected AAC files in its iTunes Music Store. The DRM protection they use is proprietary to Apple so other players will not be able to read those DRM-protected files.

Note that files that iTunes rips from CD are not protected by DRM in any way.

This post has been edited by Teqnilogik: Jun 23 2004, 20:33

[Gabriel]

Are you the creator of LAME, Gabriel?

No I am not. I am one of the regular developpers of Lame since 5 years, but Lame is the work of many people.
I am considering that Mark Taylor created Lame 3.X by giving it a strong start impulsion. The evolutions of Lame where by by numerous people, including some that spent a lot of time on it:
http://lame.sourceforge.net/developers.html

Regarding quality previsions: I am betting on AAC.
Regarding future widespread compatibility: I am betting on AAC.

[adlai]

quick question: will there ever be an AAC version of LAME? or will copyrights stand in the way?

[Busemann]

quick question: will there ever be an AAC version of LAME? or will copyrights stand in the way?

FAAC is essentially the LAME in the AAC world. But unlike LAME, it's not really up to snuff compared to the commercial encoders.

[stephanV]

I just noticed the starting date of this topic.

The more things change, the more they stay the same, no?

[gameplaya15143]

I just noticed the starting date of this topic.

lmao, i didn't even notice the date till you mentioned it

The more things change, the more they stay the same, no?

the recent ~128kbps test would suggest you are correct

I don't know how anyone else sees it... but aac seems to be having trouble living up to all the hype it was built with

[JeanLuc]

I don't know how anyone else sees it... but aac seems to be having trouble living up to all the hype it was built with

At least, when looking at the average-bitrate-to-overall-quality ratios of all contenders in Sebastian's listening test, Apple AAC comes out as the king of efficiency.

But you are right ... with no clear winner, the AAC hype might (like with most hypes) have been overestimated ...

[marcetex]

I don't know how anyone else sees it... but aac seems to be having trouble living up to all the hype it was built with

At least, when looking at the average-bitrate-to-overall-quality ratios of all contenders in Sebastian's listening test, Apple AAC comes out as the king of efficiency.

But you are right ... with no clear winner, the AAC hype might (like with most hypes) have been overestimated ...

I wonder how Dolby AAC LC @ 128 kbps would fare against LAME/iTunes AAC considering how much better their HE-implementation is compared to others @ 64 kbps.

But I have to agree that AAC (considering all it's space-aged technologies built-in) is per definition: Overhyped.

It was in development for about 8 years and clearly doesn't have any significant advantages compared to MP3:

1.
I think it's important to note that MP3pro sounds better than He-AAC @ 32 or 48 bit any day (64 kbps for dial-up is useless).

Don't believe me? Take a look at this: http://www.soundexpert.info/coders32.jsp

2.
And to sound really gruesome I will say that 16 kbps MP3Pro would hold up it's own against AAC with parametric stereo. Maybe AAC could have a small advantage at this bitrate (Speex sounds better btw., yes it's a frigging voice codec.) but the bottom-line is that they are both equally annoying at anything lower than 32 kbps. Ironically, Nero He-AAC is the most promising iteration here.

3.
MP3 LAME always sounded better at bitrates higher than 192 kbps with VBR (or even anything beyond 160 kbps) - I don't think anyone will doubt that. Let's not start talking about VBR efficiency - because AAC has none.

And If you still think or care about AAC having more potentional than MP3 - you might want to waste your time with something else until AAC3 is released.

My real point of this post is that all the AAC-iterations are rather pointless by nature:

1.
There is not one portable audio player in existance (yeah right, cellphones ) that can play He-AAC. By the time they make this happen we get 5 Gigs o' flash mem for free and/or decent ogg players that don't look like crap and fit in every pocket. (OGG Vorbis is still the best and most of all natural sounding codec @ 64kbps in my opinion).

2.
If you are serious about audio quality, you will never ever encode below 128 kbps ABR - no matter which codec you prefer. (Yes, 96 kbps AAC sounds OK but why bother?)

3.
At the end of the day it's all about the balance between audio quality/size/playback performance. If you encode MP3's at 256 kbps or higher you can as well encode in APE or FLAC because even with huge collections the difference in size usually peaks at 1-3 gigs and that's pretty laughable considering today's HDD prices. Besides, you will get the best quality possible when transcoding audio files for your DAP - no re-ripping or being stuck with a certain codec.

4.
Nope AAC still doesn't give you any real advantages for streaming audio neither and maybe, never will.


Just my take on it.

This post has been edited by marcetex: Mar 4 2006, 07:39

[[JAZ]]

I wonder how Dolby AAC LC @ 128 kbps would fare against LAME/iTunes AAC considering how much better their HE-implementation is compared to others @ 64 kbps.

I hope you mean CodingTechnologies, because Dolby doesn't have a HE-AAC encoder AFAIK. Also, Winamp had, up to version 5.2, that Dolby Encoder. (not sure if Dolby has different implementations).

I think it's important to note that MP3pro sounds better than He-AAC @ 32 or 48 bit any day (64 kbps for dial-up is useless).

We are having a test with AAC encoders at 48kbps and after that, there will be a multiformat test at the same bitrate (or at least should be, to maintain coherence).
Then, we will have a view on the subject. Soundexpert's test isn't too up-to-date (AAC has evolved in two years) and moreso, isn't as strict as those we do here.

There is not one portable audio player in existance (yeah right, cellphones  ) that can play He-AAC.

Not sure what your point is, and moreso, 2 GB and 4GB flash-based portables already exist. (ipod's at least)

If you are serious about audio quality, you will never ever encode below 128 kbps ABR - no matter which codec you prefer. (Yes, 96 kbps AAC sounds OK but why bother?)

HE-AAC is not about quality. it's about bitrate. LC-AAC isn't that different (in concept) than MP3, but at least improves on some defects it has (long/short block sizes, sfb21,...), and this is what makes it better.
It doesn't have sense to talk about future with today's parameters. As an example, people used to say that 24kbps would never be listenable, and now we have HE-AACv2.

Nope AAC still doesn't give you any real advantages for streaming audio neither and maybe, never will.

If you talk about AAC, probably you're right. If you talk about HE-AAC (vs MP3Pro), then, you're wrong. MP3pro is propietary, and there are only a few decoders outthere. On the other hand, pretty much everything can decode HE-AAC (sometimes requiring a plugin, others natively).

This post has been edited by [JAZ]: Mar 4 2006, 10:20

[Busemann]

MP3 LAME always sounded better at bitrates higher than 192 kbps with VBR (or even anything beyond 160 kbps) - I don't think anyone will doubt that. Let's not start talking about VBR efficiency - because AAC has none.

Are you really saying LAME is superior at higher bitrates without bringing proof?

[Garf]

BTW, I am quite sure that AAC at 128kbps could be way better than mp3. It just needs a few more time.

Interesting observation, coming from the LAME creator. (I might not be correct about that. I'm new at this. Are you the creator of LAME, Gabriel?)

Would the same statement apply to WMA, OGG, and MP3Pro, as well as AAC? Aren't they also variants of MP4?

No. Though, OGG fixes several deficiencies in MP3 in very much the same manner as AAC does. We don't really know what WMA/WMAPro do, but I expect it to be the same.

MP3Pro is just MP3 with SBR extension. It's still flawed in the same way, and not even an open standard.

[gameplaya15143]

1.
I think it's important to note that MP3pro sounds better than He-AAC @ 32 or 48 bit any day (64 kbps for dial-up is useless).

Don't believe me? Take a look at this: http://www.soundexpert.info/coders32.jsp

2.
And to sound really gruesome I will say that 16 kbps MP3Pro would hold up it's own against AAC with parametric stereo. Maybe AAC could have a small advantage at this bitrate (Speex sounds better btw., yes it's a frigging voice codec.) but the bottom-line is that they are both equally annoying at anything lower than 32 kbps. Ironically, Nero He-AAC is the most promising iteration here.

listening tests have shown mp3 and lc-aac to be pretty much on par... so mp3pro and he-aac should sound about the same at the same bitrates, since the only thing making them sound 'listenable' at low bitrates is pretty much the same SBR (i recall mp3pro's sbr and CT's aacPlus SBR sounding about the same to me)

so yea, the only real difference is that one has open standards (and lucky for us its the one with *more* potential supposedly )

[marcetex]

I wonder how Dolby AAC LC @ 128 kbps would fare against LAME/iTunes AAC considering how much better their HE-implementation is compared to others @ 64 kbps.

I hope you mean CodingTechnologies, because Dolby doesn't have a HE-AAC encoder AFAIK. Also, Winamp had, up to version 5.2, that Dolby Encoder. (not sure if Dolby has different implementations).

This post has been edited by marcetex: Mar 14 2006, 17:29

[halb27]

...

3.
At the end of the day it's all about the balance between audio quality/size/playback performance. If you encode MP3's at 256 kbps or higher you can as well encode in APE or FLAC because even with huge collections the difference in size usually peaks at 1-3 gigs and that's pretty laughable considering today's HDD prices. Besides, you will get the best quality possible when transcoding audio files for your DAP - no re-ripping or being stuck with a certain codec.

It's true: for storage on PC, notebook or specialized hifi equipment it's best to go lossless.
But mobile DAPs are widely used, and going lossless is not very promising in the very near future. This is not so much a question of storage space (which is not very restrictive even today) but of the time until the battery is down.
With a large file size HD is turned on often which has a very negative impact on playing time.

I originally wanted to use my iRiver HP140 using very high bitrate wavPack lossy but gave it up in favor of high bitrate mp3.
mp3 is in fact a power saving format, and I'm happy with playing time @ 270 kbps mp3.
As for quality I think you're wrong. I'm just doing a problem sample listening test (I'll report on that), and you need 192 to 224 kbps with mp3 to get satisfying results. You can get the same quality level at 160 kbps with Nero6 AAC or Vorbis aoTuv b4.
Using 128 kbps on these samples yields total crap with mp3 and is a lot better with AAC or Vorbis. AAC and Vorbis results at rather low bitrate are more homogenous throughout different tracks.

Anyway I personally prefer mp3, but only because I don't care about the higher bitrate I have to use knowing battery life is still very good. And I like universal usage.
If AAC were widely supported and if it doesn't drain on battery life too much (don't know nothing about that) I would prefer AAC. Vorbis is widely supported but unfortunately is a big battery drainer (at least with my H140).

[Serge Smirnoff]

,Mar 4 2006, 12:10 PM]

I think it's important to note that MP3pro sounds better than He-AAC @ 32 or 48 bit any day (64 kbps for dial-up is useless).

Don't believe me? Take a look at this: http://www.soundexpert.info/coders32.jsp

We are having a test with AAC encoders at 48kbps and after that, there will be a multiformat test at the same bitrate (or at least should be, to maintain coherence).
Then, we will have a view on the subject. Soundexpert's test isn't too up-to-date (AAC has evolved in two years)......

Yes, I confirm - 32/48 tests are not up-to-date and its clearly seen from the versions of codecs used.

,Mar 4 2006, 12:10 PM].... and moreso, isn't as strict as those we do here.

Indeed, below 128 kbit/s the only difference with HA tests is the absence of disclosed reference sample which is pretty useless in 32,48 kbit/s tests. The more serious problem with SE tests is lack of participation, so the results are not so reliable as HA ones.

This post has been edited by Serge Smirnoff: Mar 18 2006, 12:50

[Ivan Dimkovic]

I think it's important to note that MP3pro sounds better than He-AAC @ 32 or 48 bit any day (64 kbps for dial-up is useless).

Don't believe me? Take a look at this: http://www.soundexpert.info/coders32.jsp

This is quite incorrect.



This is the 32 kbps test done with 2 years old encoder from Nero, which is quite lower in performance than today's (used in current Gabriel's 48 kbps AAC listening test) - with the current state of the art, I'd expect that HE-AAC scores quite higher than mp3Pro, which will be tested very soon on public 48 kbps listening test.

Then, there are results from CT, who invented SBR used in both mp3Pro and HE-AAC:

http://www.ebu.ch/en/technical/trev/trev_291-dietz.pdf

Figure 9, on average at 48 kbps, then current HE-AAC implementation scored better than mp3Pro.

It is important to note that HE-AAC is an international standard, open to everyone to implement - while mp3Pro is a single-vendor proprietary solution.

This post has been edited by Ivan Dimkovic: Mar 18 2006, 13:22

[IgorC]

It is important to note that HE-AAC is an international standard, open to everyone to implement  - while mp3Pro is a single-vendor proprietary solution.

As conclusion HE-AAC has more hardware compability and will have even more.

[cosmin]

Okay, back a month or so ago when Roberto did his listening test @ 128 kbps for all the formats LAME and iTunes essentially tied.  Can someone please give me a quick explanation how LAME, based on the MPEG-3 standard somehow tied with iTunes which is based on the MPEG-4 standard?

Am I missing something here?  I would think the newer MPEG standard would crush the bejeezus out of its predecessor.  What's the deal?

It appears to me that you are missing the following fact:

That particular comparison was done between 128k AAC CBR versus 128k MP3 VBR. At that time, iTunes was only able to encode AAC in CBR. (I don't remember very well, perhaps other AAC encoders were VBR, but in that case I would assume they were of inferior quality.)

A fair comparison must be done between a good 128k MP3 VBR encoder (e.g. LAME) and a good 128k AAC VBR encoder (e.g. iTunes v6). Check out the most recent listening test to see how it goes:
http://www.maresweb.de/listening-tests/mf-128-1/results.htm

Best regards,
Cosmin