I've been experimenting with the iTunes and Nero aac encoders to get an idea of how well this technology works and so far I'm a little underwhelmed. It doesn't seem to achieve transparency at significantly lower bitrates than lame -aps. In fact, the Nero transparent setting seems to encode at slightly higher bitrates than lame -aps on the material I've tried.

There is some portable support, but certainly not as extensive as support is for mp3. Furthermore, the Nero encoded files I've made crash my ipod. Vorbis portable support seems to be just about as good, although I expect AAC will probably bypass Vorbis in this respect eventually.

The high-quality encoders are very difficult to automate and don't seem to be available in command-line forms at all (nencode hack works I guess), espcially not cross platform. Ogg and mpc are much more flexible in this regard.

Consistent gapless support doesn't seem to be available either.

The only real advantage I can see at this point is that AAC sounds marginally better than mp3 at lower bitrates, but I rarely fill up my 10 gig ipod as it is so that doesn't appeal to me very much.

Am I missing something? Right now I'm inclined to stick to mpc and keep transcoding to mp3 for my ipod, especially now that mpc seems to be showing renewed signs of life.

Good synopsis imho.
HE-AAC and other ways to bring compression down to 1:20 would be really interesting for cheap portables with flashmem, but there is no hardware for it.

If you want to stay at todays ~1:7 "transparent" rates, I see no reason to go away from LAME. I am happy that mp3 is such a strong standard, so that all hardware and software solutions HAVE to support it, in addition to their various prop. formats that come and go

AAC is technically better than MP3. It has the potential in the future to bypass LAME MP3 and become transparent at lower bit rates. As it is rather new compared to the decade old MP3 technology AAC has yet to be fine-tuned for transparency. Nero's AAC encoder is still very young and Apple's AAC encoder only has CBR support. Apple's AAC encoder performs exceptionally well in my opinion considering it only uses CBR and it's completely free. I use iTunes AAC at 128kbps. It suits me just fine for almost all of my music and the artifacts I notice rarely are not annoying to me like MP3's artifacts are. In the future I see AAC getting more and more support and eventually matching MP3's popularity in hardware and software players. I've decided to buy an iPod so I've switched to AAC. Foobar2000 and Winamp now support AAC natively it so you have two great player choices on the PC platform. Though, I understand what you are saying. LAME MP3 using alt-preset standard pretty much guarantees transparency on any sample while that isn't true with AAC encoders today. Though I will add when I first heard a 128kbps AAC file it blew me away at how good the quality was compared to a 128kbps MP3. AAC development has picked up tremendously recently so I expect great things to come of AAC in the future. Now that iTunes is for PC I see AAC growing even more popular among the public where before it was pretty much a Mac-only format since when iTunes was Mac-only software only Mac users had the ease-of-use iTunes application to encode AAC files. Who knows, someday FAAC may also have built-in presets and become the LAME of AAC encoding.

Well, if you just compare something like Lame --preset standard to similar MPEG4-AAC, it's true that there isn't much difference for most listeners.
As a format here's why IMO MPEG4-AAC is better.

- Lower license fees and no streaming fees
- Generally better quality (depends on the implementation, but at least the potential is there)
- Standard covers the latest technology and is updated as better technology appears (SBR, Parametric Stereo, Multichannel, Spatial Audio coming etc. etc.)
- Possibility for clearly better quality than mp3, the lower the bitrate (of course the higher the bitrate, the difference gets smaller)
- Lots of development and exciting new technologies appear quite often

Now that AAC-players start to be the standard even in the new Nokia mobiles and more hardware support appears every day, I believe this year is going to be even bigger breakthrough for MPEG4 than last year was. I except especially streaming to become a bigger sector.

But still, if you just want to play your Lame -preset standard files, I guess MPEG-4 AAC doesn't bring anything revolutionary for you (practically nothing will).
Personally I think MP3 starts to be quite boring as a technology.

well, when you consider that it is an extension/continuation of psytel aac - it doesn't seem all that new at all (around for ~4 or 5 years )


later

One thing to consider here, is future development. MP3 as a technology is now quite old, and it has been tuned to work very well, especially LAME. But, there is probably little left to improve. It has very likely reached the maximum quality vs. size that is possible given the limitations in the encoding format. And it still has "issues" with some hard to encode samples.

AAC/MP4 on the other hand is quite new, and is being heavily developed by many companies and individuals. It improves constantly, and hasn't, IMHO, reached the same level of "tuned-ness" as MP3. There is still room for improvement, but it is already sounding quite good.

I'm ripping all my music first to FLAC, then to iTunes-encoded AAC at 192kbps for my iPod. To my ears, AAC at 192 comes very close to LAME aps quality at lower bitrates with my music. I'm not listening using $500 high-end cans, but I do have decent Sennheiser's for my iPod. I suppose that in a quiet listening environment with really good cans I could ABX LAME aps vs. AAC 192, but only on a good day. But, that's not how I listen to my iPod, I'm usually outdoors where I would never hear the difference anyway. At home, I listen to the FLAC files streamed off my hard drive.

I was encoding everything in MPC SV7 insane, but after a few false starts and having to re-rip some CD's all over again, decided lossless was the way to go. If I want to use a new or different format in the future, I'll just decode the FLACs to WAV and re-encode, no loss in quality. Transcoding from MPC insane just didn't quite cut it with some music.

I agree it would be nice to have commandline encoding, especially for iTunes/QT. I have a complex ripping routine right now, which also overcomes some inherent problems with the iPod, like poor volume leveling support:

1. rip to FLAC in EAC, keep the WAV's
2. ReplayGain the WAV's using WaveGain
3. drop the WAV's in iTunes, tag, encode

It works, and I've recently completed my mass encoding project of over 350 CD's this way.

Most of these are 'implementation issues' rather than issues of the format (MP4/AAC) itself. Apple's noncompliance to the standard isn't the standards fault.
Using Nero and foobar2000 you get perfectly gapless playback and flexible tagging now, it's only a matter of time until other implementations catch up and implement more parts of the standard.

AAC's theoretical advantage is proven by FAAC, which even though it lacks almost any kind of psychoaccoustic model, performed reasonably well in recent tests. (its performance left Gabriel quite suprised)

Also you won't see much development happen to MP3 anymore (there hasn' really been any during the last 2 years) and LAME (since 3.90) is considered to be an "as good as it gets" implementation for bitrates > 96kbps (LAME4 might prove us wrong here), while MP4/AAC implementations improve almost on a daily basis (just look at Ivan's recent post about PS).
Overall right now MP4/AAC might be equivalent feature- and performance wise to LAME MP3 at transparent bitrates and for the sake of compatibility I'm still using MP3 for some things too, but disregarding it as 'uninteresting' like you did is a bit short sighted in my oppinion

dev0

May I remind you of our beloved TOS#8 here?
Please don't make such statement without supporting them with hard facts. Rjamorim's multiformat test last year has proven Apple's AAC implementation to perform significantly better than LAME MP3 at 128kbps.

Yeah you're right. I forgot Nero's AAC encoder was just a continuation of the Psytel AAC encoder so I guess it wouldn't be that young afterall. My bad. Thanks for the correction

I'm aware of all the potential advantages of AAC as a technology but it seems like there aren't many concrete advantages at present. AAC seems the technlogy best positioned to supersede MP3 (aside from WMA). I understand why all the compression geeks (like me) on this board are attracted to it. I just don't see any real advantage at present for using AAC now and I see several disadvantages. I expect this will change but right now I don't see why I'd bother if all I want is good sounding and reasonably compact audio files. I guess this is why I ripped all my CDs as single-file flacs.

What I'd really like to see is FAAC develop to the relative quality level of LAME vs the other mp3 encoders. I'd settle for a command line Nero encoder that works on Linux though.

Yes, the disadvantages of using AAC are hardware and software support mainly and there are minimal free implementations. Of course every new technology is going to need time to gain acceptance and compliance with hardware and software. The thing is, if people don't use the new technology it will never be implemented in new hardware or software. There has to be a demand for it. If you decide not to use AAC simply on the fact that MP3 is more popular then you are giving hardware manufacturers and software developers no reason to include support for the new technology as most of its user database won't be using it. So that's why I try to embrace new technology and get other people to try it. By doing that you will spread the word about AAC/MPC/Ogg Vorbis/whatever your format of choice is and get other people to start using that better format. Eventually then hardware and software developers will see the growing demand for the new technology and start implementing support for it.

I too would like to see FAAC grow and improve in quality. I would like to throw out a suggestion to FAAC developers: make your encoder easier to use. Rather than using quantinization values allow the user to enter in a bitrate they want to encode at or develop some presets like LAME has and tagging support would be great (iTunes compatible tagging since that seems to be what everybody is using since there is no tagging standard defined in MPEG-4).

Edit: I just visited Rarewares and saw FAAC 1.24 beta was released. I read the changelog and discovered that it now has iTunes-compatible tagging support from the command line. So no need for that feature anymore.

6 years.

http://pessoal.onda.com.br/rjamorim/rrw/aacenc.html

are there any threads/samples where lame is much better than nero-aac for example?

I use QT/iTunes 192 & 224 kb/s, not because I think it sounds substancially better than LAME -aps or -apx, but because even with using the "fast" tag, LAME is still ~5X slower encoding on my machine (dual 450 G4).

If you use FLAC or another lossless or uncompressed format for most of your home listening, portable lossy encoding time can matter a great deal when chosing an encoder. Since any faster MP3 encoder is usually thought of generally inferior to LAME, and iTunes AAC is at least as good, I go the fast route - AAC.

i stand corrected! all hail teh uber archiver!

...really tho, when was the first alpha released?

i knew of 1.0/2000 (~4yrs) and 1.0alphas/1999 (5yrs) but can't find anything released in 1998?

If I could automate the flac -> aac conversion and the resulting aac files would play on my ipod I'd probably do this too. Unfortunately neither is the case right now.

Lame 3.96beta does seem to be quite a bit faster than 3.90.3 was.

You mean the standard is going to be updated with new tools as they are invented? Doesn't that rise severe compatibility issues (thinking of hardware devices)?

Which tools does a device have to support to be fully HE-AAC compliant?

Side question: Is there something like Parametric Multichannel?

Not that much,

- Improvements are always backwards-compatible, i.e. old decoders still could decode part of the bitstream maybe with lower quality, but still with some result

- Standard is improved only when there is significant potential compared to the state of the art, and the improvements themselves are made in such a way that there is a plenty of room for future optimization without changing the bitstream syntax.



LC AAC and SBR



Not yet, but it is planned, bringing 5.1 audio to 48 kb/s

Only LC and SBR for HE compliance? What about PS or is this already included in SBR?

Is there any chance that LTP will ever be able to improve the quality? Did it come out to be too complex to be properly used or is it just not usefull at all?

PS is not required by SBR at the moment, and I do not have the final document (not yet finished) - right now it is "Ammendment 2: Parametric Coding" - however it might, as a tool, be put as required for some profiles.



It is able to improve quality, but not that much - considering the complexity imposed on the codec, I'd skip it

Thanks, Ivan!

Your patience and your knowledge are always appreciated.

Don't forget that Nero AAC is quite young if you count how long it has been professionally developed.
Ivan started developing psytel aacenc at the age of something like 16, and it was more like a hobby before Ahead hired him soon 2 years ago (iirc). So in that sense Nero AAC is young.

Well, from that point of view, Lame is even younger, because it has never been professionally developed. The same can be probably said about Musepack.

I believe that what really counts is the effort put into development, and not how long it spent being developed, professionally or not.

No. Lame MP3 is not a professinal product, rather an open source project which has had many active and talented developers during many years. In no way Lame or even less mp3 can be said to be very young anymore.
Musepack is an order of magnitude easier code to control simply because it doesn't have block switching issues, and big part of the DSP in it is directly derived from mp2.

You can't possibly know how much effort and how serious went to some codec during some time, if you aren't the developer yourself.

Sorry for the ignorence, but what does "Transparency" mean?

Transparency means that the compressed audio is indistinguishable from the original CD quality audio.

It's really a testament to the quality of mpc that even after such a long period of dormancy musepack still stands up so well to the other codecs in terms of compression speed and quality vs bitrate. My mpc files are pretty consistently 20% smaller than my lame -aps or nero -transparent files and are very quick to encode.

I am 110% with Kuniko here... All the 'progress' I see nowadays is tuning within the 'low' bitrate range. I'm gonna be fairly bold here, and state that most of us people reading / typingto this forum are more interested in high bitrate, transparent bits of music. We do not care at all about this "YUO CAN GET 300,000,00 SONGS ON A CD" style stuff thrown at us (nor for Christine Aqg.....)

There is little concrete to distinguish a decent --aps from any aac.

YES, technically AAC is more advanced. Lets hear it, hardcore tuned up, like Dibrom did with Lame.

scratch this

I would say lower bitrates are being tuned more because with newer codecs it is possible to achieve transparency at a lower bitrate rather than needing 200+kbps like MP3. I realize that AAC has not been as fine tuned as LAME MP3 and that a --aps MP3 will most likely be more transparent than the majority of AAC's at a similar bitrate at the moment. But if fine tuned as well as LAME MP3, AAC can reach transparency equal to --aps at lower bitrates. Of course it's going to take a lot of time before we see this happening. But at the moment I am satisfied with iTunes AAC at 128kbps. With the development going on with the Nero AAC codec I think we'll see it catch up to LAME MP3 --aps quality sooner than later.

my AAC files have a fair amount of both transparent encodings (usually around 128-130kbps for most current CDs) and very likely more in the 40-60kbps range for my website once I really get started on those.

Having said that, I would agree that most people here are far more interested in the LAME -aps range/quality.

Continuing on Lev's thoughts, there's so much talk about how aac lacks many of mp3s limitations and is so much more modern, yet i haven't seen been able to find much evidence that anyone has taken advantage of the headroom they keep telling us it has. instead of knowing that aac can do what -aps does with a lower bitrate, i'd be interested in seeing how transparent it is above -aps and where it hits a law of diminishing returns by going higher, essentially its -aps/-ape divide. if i knew that aac at, say 224 or nero's extreme reduced the TOTAL number of problem samples i'd find by a signifigant rough percentage vs. -aps, that would be hard to ignore. there's still a huge gap in size and usability between "good" mp3 and aac files(160/92/-aps) and going lossless, i think a lot of people wouldn't mind seeing a few planks laid towards bridging it

...and what's up with aac tagging? id3 has SO MANY glaring problems, but itunes aac tags have FEWER fields than id3v2.3? and the bpm/discxofy business is so useless to most people that it practically doesn't even count. no moods, keywords, guest artists, track artist vs album artist, etc

You must keep in mind iTunes tags aren't standardized tags. When MPEG wants a standard, they don't come up with a half-baked solution like what ID3, Apple and Xiph do. They go with full blown MPEG7 - thousands and thousands of tags for your pleasure.

I can understand why there's so much commercial interest in low bitrate encodings. Smaller files are cheaper and faster to download and they're less of a threat to CD sales. They just don't interest me much because my goal is getting all of my music on a 250gig drive in a format that I can't distinguish from CD. AAC, MPC, Ogg and MP3 all work pretty well for this but MPC still does it best.

i read somewhere that they already did standardise tags for mp4 in mpeg-7 (dunno if it is finished tough), basically by using a "meta" atom (not placed in udta (user data), like itunes)

still as expected it will surely not be compatible to itunes tagging

Wait, so you have one of the few portables that supports AAC, but you're archiving in MPC and re-encoding to MP3 for the portable?

That's part of my frustration. The aac files I've encoded with Nero crash my ipod and there's no way to automate flac -> quicktime aac conversion.

All my stuff is archived on DVDs in flac files but I'll usually just transcode mpc to mp3 because it's easier than pulling out the disks.

What about this solution? http://www.dbpoweramp.net/codecs/dBpowerAMP-codec-QT-mp4.exe I tried it to convert flac to quicktime aac. It works, albeit at a leisurely pace.

I think the Openess of Vorbis is a good enough reason to move over to only that format and to start loads of petititions to get companies like Sony, Apple, Phillips, Warner etc to start supporting it too in both hardware and software.

Right now Dolby, Fraunhofer/Thomson and MS are probarly doing all they can to prevent companies to get in touch with Vorbis, don't you think? It is as abvious as that the theory about Adobe having some deal with Microsoft and Apple about not making Linux versions of any of their products.

well, You can't realy say that the development over MP3 has stopped.

Check this out, now it's down but check it a few days later, you'll be surprised.

WTF does vorbis has to do with this thread???

You are polluting this discussion. If I were you I'd stop spreading unproven BS and blindly advertising vorbis on unrelated threads as you already did before here.

I think the clear conclusion is that while AAC may have much more potential than mp3 & would be a more interesting format from a developers point of view but if you are a user who wants the highest audio quality possible in the 160-320 kbps range you are better off with a format such as lame mp3 or Musepack.

Although neither of these format have been commercially developed listening tests prove that some very talented people have been in involved their development.

Yeah, I think MPC really embaraces the compatition. I would like to see a few other features for SV8, and some additional tuning at lower bitrates (for portables). One of its main stengths that has not been mentioned here is the low CPU overhead for decoding, which on a portable would translate into longer battery life. All of the NEW technology requires alot more CPU power, with little benefit so far.