Hello,

seeing that the last update to Nero AAC was made almost a year ago (2008.09.17 - v 1.3.3.0) I wonder if its still being actively developed. If so, can some of the devs tell
1) if we should expect an update to it this year still?
2) what parts of the codec are being worked on?

Anyway, I should test 1.3.3.0 for multichannel coding because it sports improvements in that area, while the last version I tested was 1.1.34.2, which performed worse then two other codecs tested.

Regards

Yes, it is being developed. All parts of it.
Version 1.3.3.0 should indeed contain improvements of quality for multichannel.

Thanks for the response. Could you also comment on question 1) ?

Why do you choose to have such long periods between releases?

Why do you need more releases and faster development? Did you detect problems that have to be solved as fast as possible?

One reason might be that the people who work on the Nero AAC codec (muaddib and Menno only AFAIK) are involved in other tasks as well. Tuning also takes a lot of time especially for higher bitrates where only one or two people report problems.

It isn't any longer than other lossy encoder updates. Lame 3.97 was released in August 2007 (someone please correct me if I am wrong) and Lame 3.98 was released in July 2008. That was a year in between the new versions. Nero 1.3.3 came out in September of 2008. So one would believe that a new version might come out around August/September of 2009. However, a new version might not be needed. What is the point of releasing a new lossy encoder if it doesn't really improve any type of functionality? Releasing a new encoder for the sake of releasing a new encoder would be a good way to get people to stop using it.

According to http://lame.cvs.sourceforge.net/viewvc/*ch...ml/history.html :

LAME 3.96 April 11 2004
LAME 3.97 September 24 2006
LAME 3.98 July 4 2008

I'm asking because I'll be doing plenty of encodes, so if a new release with quality improvements is imminent of course I'd wait. Speed improvements would also be interesting, though that would be just a bonus.


According to muaddib all aspects of the codec are being worked on, so I'd expect improvements in all areas.
Unless it's a single big new feature that would take 1 man year+ to complete I don't see why hold back such improvements for so long. If it's new experimental tunings, why not ask for feedback from the community? That can speed things up compared to doing everything in house. You mention Lame as example for long periods between releases, yet there are plenty of beta builds for exactly this purpose.


If fixing reported issues is all that's being done I'd call it "passive development"...

Thank you for correcting me. I couldn't readily remember 3.97's release. So there was a two year break in between each version.



That is understandable. However, ripping to a lossless archive would eliminate any hassle. Sure, your computer would be occupied for a weekend or two during the encoding process (ie lossless to lossy) but that is a lot better than having to rip every CD. I could see being worried if you were ripping your CDs straight to Nero AAC but then again, the prices of storage have come down so much that there really is no excuse for not having a lossless backup of content.



The issue is that the developers are likely involved with other projects and other aspects of Nero AAC development. Nero has a whole fleet of applications that range from CD/DVD burning, video encoding, video editing, and so on. I am sure that some of Nero AAC's developers are involved in those aspects. Much more than passive development in terms of the big picture.

Well, except for fixing reported issues, what else do you expect from an encoder that has to respect certain ISO standards? There isn't much left to implement as new feature.

Being a long time HA member (though more lurker then poster) I know this would be the ideal way.


...unless one has a *really* large collection


Yes, very likely AAC isn't their priority at all. Thus my OP question.


Well, look at some other codec projects, x264 for instance. They're also constrained by the h.264 specs. Yet subjective quality tuning can be refined further and further. Ofc someone could argue that those large rooms for improvement come from the less well understood/modeled psy-visuals compared to psy-acoustics, but lets not go there Let me rephrase as a question: Do you think AACs potential is near exhausted by current encoders?

I would venture to say that if one had a "really" large collection of CDs then one should be able to afford a "really" large hard drive(s).

The cost of hard drives is below $0.05 to store a typical lossless-encoded album. The cost of the CD was orders of magnitude larger.

Just let me get something straight since maybe I didn't understand you correctly. Do you consider tuning = passive development? If not, then muaddib already said that Nero AAC is being worked on beyond the point of bugfixing only. If yes, then x264 is also only being passively developed since we're talking about tunings and not brand new features like the introduction of PS in Nero AAC some time ago for example.

As to your question - AAC is a broad term. You have LC-AAC, HE-AAC, HE-AAC with PS, SLS, etc. Then you have multichannel, stereo, mono. You have all kinds of sampling rates. All these can be tuned but this of course takes time and not much was tested in some of these areas. It also comes down to what people want in first place when you have limited resources.

That was pretty much my point. Not to go too far off-topic but the cost of storage has come down a lot. Even "name brand" PC manufacturers are releasing inexpensive storage solutions. HP is selling a 1TB USB 2.0 hard drive for $100, internal 1.5 TB hard drives are the same price, and 2TB drives are $200. Someone could easily backup thousands of CDs on a couple of 2TB drives in a lossless format (ie FLAC, WavPack, ALAC, etc.). A wise investment considering the initial cost of CDs.



I am not saying that their AAC encoder isn't a priority, it is just part of other projects on the developers' plates. That still doesn't explain your "passive development" comment.

For 2-channel audio and bit rates beyond 128kbps: yes, most likely. We might see improvements for a few critical items, but overall, possible improvements are minimal.

Chris

Passive refers to your description of the development approach (only fixing reported problems), which is somewhat different from what muaddib implied.


True. And I should have said I'm interested in transparency bitrates, mainly stereo but also some multichannel, so LC.


Alright.


If its a matter of lack of testing, then my suggestion to have the community give feedback on dev builds still stands.


So does Nero listen to user wishes?

There is still room for improvements for Nero encoder IMHO. http://www.hydrogenaudio.org/forums/index....showtopic=66949
Probably even at high bitrates.
Afaik Apple is preparing a new encoder (at least SBR part). http://www.hydrogenaudio.org/forums/index....c=71835&hl=

'Never say Never' is still correct phrase.

Look at LAME. If 10 years ago you'd ask someone if that was possible to achieve good quality with MP3 at 128 kbit/s. Nobody would believe. Today http://www.listening-tests.info/

Another sample is already mentioned x264. H.264 is supposedly 2x better than MPEG-2 by standard ISO definitions. It's incredibly how x264's devs (Dark Shikari, Loren and other not less important developers) managed to obtained even better results.

I could obtained the twice better results than Xvid (at bitrates where x264 was already enough good). With last alpha revision:
http://forum.doom9.org/showthread.php?t=148686

MSU also informed near twice better compression ratio for older revision of x264 comparing to Xvid.
http://compression.ru/video/codec_comparis...64_2009_en.html
http://compression.ru/video/codec_comparis...arison_2009.pdf
See the pages N23 and 26 Table8 and 10 respectively : SSIM (more realistic metric than OPSNR)

good to know that it's always in development,Nero is supposed to be the "Lame" in the AAC World IMO. At least they offer us the encoder without having to stay tied up to a particular software (ie iTunes and Apple AAC encoder [on Windows,XLD is a lot better on Mac]). Can't wait for further updates! Keep up the good work!

The only request I have for the Nero Developers is to please somehow make use of modern multi-core processors. The Audio Quality is superb and I have no programming skills, but x264 and other video codecs have been well optimized to take advantage of 8 or more cores yet when I encode audio files my CPU is barely utilized. I don't have a new CPU either it is an athlon x2 3800+!

I love the audio quality provided by this codec!!! I am only pointing out one area in which all modern audio codecs could use some more development.

Thanks to the nero AAC Developers for their hard work!!!

I don't know much about AAC or lossily audio compression in general but my guess from what I do know is multithreading audio codecs is a lot more difficult then video codecs due to vast difference in way they work (there's vastly more room for parallelism in video encoding).

There's also the question about 'what for' I guess. Compressing audio is already so fast there seems limited point. Perhaps if you're compressing 10 hours of audio but that's probably fairly rare. Far more likely, the only time it's likely to take a while is if you're doing a lot of smaller files at once in which case file level parallelism makes a lot more sense and is already easily done and doesn't require the work of the Nero developers.

A discussion was raised not too long ago asking for multi-threaded lossy/lossless encoders. From what I remember, there is basically no point especially whenever two (or more) encoding instances can be started. In other words, I can encode two tracks at the same time instead of focusing both cores/threads on one song. Sure, the one song can be encoded a little faster but I would rather wait the extra few seconds and get two songs out of the process. I know dBpowerAMP takes advantage of multi-core/hyperthreaded processors as I can have two encoding streams occurring at the same time (without issues) on my AMD Athlon X2 powered desktop and on my Intel Atom N280 powered netbook (the former contains two core and the later is hyperthreaded).

kornchild2002 was referring to this discussion (I suppose).

As far as I can see multithreading wouldn't lead to gain when CPU is bottleneck, but when hard drive speed is bottleneck possibly there's a encoding speed gain. With todays mainstream PC's encoding 1 file per core or hyperthread is the fastest solution (with pipelining).

Wow.

The suspense is killing me.

In another thread, one member said it might be out in weeks, or even days.

Please let us know when you determine the date.