Time for a new AAC guide?

[Kent Wang]

HE-AAC is being released, lots of things have changed since June 2002.

[hans-jürgen]

Where is the old guide? ;) For a start you might want to read the Wiki on http://www.audiocoding.com/wiki/ which is a user knowledge base about all audio formats, also AAC and MP4.

[bond]

http://www.doom9.org/mp4.htm

perhaps there should be nero and quicktime mentioned too (and that aacmachine has problems with multichannel)
also mentioning the 3ivx splitter would be great...

[hans-jürgen]

Originally posted by bond
http://www.doom9.org/mp4.htm I wasn't sure if Kent meant that one, because it had been updated in May 2003. Anyway, in the Sticky thread about important links in the "New Formats" board I tried to list all available sites with current information about creating AAC/MP4 files, also the link above.

perhaps there should be nero and quicktime mentioned too (and that aacmachine has problems with multichannel)
also mentioning the 3ivx splitter would be great... There are still some flaws in the MP4 guide, one being the recommendation of PsyTEL/AACmachine for multichannel audio. Another is using the -h switch in mp4creator, because you don't need to hint video and audio tracks for local playback of MP4 files, only for streaming them via RTP. This will only blow up the file size unnecessarily, because it produces separate time stamp tracks for the elementary streams in the container.

And of course I don't like the generalizing comment about FAAC and FAAD2 that "could be better", because this is true for all codecs all the time, especially for PsyTEL in a multichannel encoding because of its M/S matrix bug. Last but not least FAAC is the only AAC codec at the moment that enables direct transcoding from multichannel AC-3 files without temporary WAVs (splitted or not) in between, and also at a much higher speed than PsyTEL, by the way. As far as I can see, it will take some time before QuickTime or Nero will provide this option. And why there should be no way to play multichannel AAC and/or MP4 files yet I don't know either... maybe only with those open source codecs... :rolleyes:

[Doom9]

hmm.. I haven't completely read up about it so my info might not be complete, but there shouldn't be any problems on the decoding side with coreaac and the next 3ivx release, the splitter and playback filters are already available today, but that leaves an encoder outside of Nero missing. Is there a way to pipe decoded AC3 data to an AAC encoder without intermediary files now?

[hans-jürgen]

Originally posted by Doom9
hmm.. I haven't completely read up about it so my info might not be complete, but there shouldn't be any problems on the decoding side with coreaac and the next 3ivx release, the splitter and playback filters are already available today, but that leaves an encoder outside of Nero missing. Is there a way to pipe decoded AC3 data to an AAC encoder without intermediary files now? The answer to both questions is FAAC (combined with foobar2000), and you can find a description how to do this on the following Wiki page and in several threads in this forum, e.g. in the "Audio Encoding" board, too:

http://www.audiocoding.com/wiki/index.php?page=FAAC

If you combine my username with "FAAC", the search function should list all relevant threads. Maybe I will update the Sticky thread with Q&A about new formats again, so there's no further excuse... ;) Sorry to sound like ranting, but I've read so much BS in the last few weeks about FAAC and how bad it would be (I don't mean your MP4 guide here) that it's simply enough now. :)

[JohnV]

Originally posted by hans-jürgen
Sorry to sound like ranting, but I've read so much BS in the last few weeks about FAAC and how bad it would be (I don't mean your MP4 guide here) that it's simply enough now. :) Ermm.. Even the developers of FAAC admit its inferiority in quality, so it's a bit questional which claims are actually BS and which not.. ;)

Ivan D. tested the latest build and said it's not much better (true Ivan may not be the most objective here, but it was a private conversation). However, some people are now afraid to say some facts to you in public about FAAC, because you are otherwise very knowledgeable about MPEG4, and they don't want to see you leave various boards.. (though I know you left HA)

[Doom9]

hmm... so what's the best encoder then, from a quality standpoint? Nero?

[JohnV]

Originally posted by Doom9
hmm... so what's the best encoder then, from a quality standpoint? Nero? I thought there was a rule which prohibited asking what is the best.. :D

Seriously, as always, there's no simple answer to that.. For CBR - it seems to be QT AAC 6.3, for average bitrate VBR at medium bitrates - probably Nero, for lowbitrate Nero HE AAC, for multichannel again Nero (at least until QT 6.4 comes with multichannel suppot), and so on..

One thing is clear though, FAAC is clearly worse than others.. Practically no-one (devs, testers, bling listening test results) else except Hans-Jürgen denies that... :)

[Kent Wang]

Actually, will HE AAC make an impact on movie-encoding? It's optimized for low bitrates and not high quality, but is the quality provided at its optimal bitrates on par with, for example, MP3 VBR 160kbps? No matter how efficient it is, if it can't generate quality above that level I'd stick to using Ogg because most of my rips are 2CD's and I can afford the extra bits to give to the soundtrack.

[hans-jürgen]

Originally posted by Doom9
hmm... so what's the best encoder then, from a quality standpoint? Nero? If you mean sound quality here, my advice is always to trust your own ears, not others (neither mine nor those self-proclaimed "golden guru audiophile" ones). By the way, I never wrote that FAAC would be the best-sounding AAC codec ever, in fact I described the remaining artifacts and suggestions how to solve these problems in the Audiocoding Wiki and forum some time ago. That's why I'm quite sure that I know how the different versions of FAAC sound, and with the default setting the latest v1.18 is competitive to other formats like Vorbis, MP3 and even Musepack with some samples (not with others, that's also true). By the way, there are at least two users of this forum who actually tried FAAC for video ripping (not with test samples) and liked it, so it can't be that bad.

Nero is probably the best AAC codec at very low bitrates now, because it's the first with a HE AAC implementation. If it's better than QuickTime at higher bitrates is still an open question, because the last listening test restricted its performance by forcing it to use CBR instead of the better VBR preset and favoring QuickTime's CBR-only encoding method that way.

Coming back to my first remark, in my opinion there are other things to consider in a codec's performance than top-notch audio quality with critical test samples, too, especially when using it for the purpose this forum is all about. And the features of FAAC make things much easier than e.g. QuickTime (no batch encoding, no multichannel support, no direct transcoding from AC-3, no VBR encoding = waste of bits for less critical details in the LFE or Centre channels of a future 5.1 implementation). Nero has less problems in this "usability" aspect, but a direct transcoding from AC-3 is still not possible, as far as I know. This might change either with an official AC-3 support or with an input plugin from Mausau though.

Furthermore it might be considered to be a bit "bloated" to download and install either the full Nero or QuickTime package just to convert your AC-3 files to AAC/MP4, while FAAC/foobar2000 is only ~1.6 MB small (full package from Case). And last but not least it's free while the other two apps are not.

[JohnV]

Originally posted by hans-jürgen
If you mean sound quality here, my advice is always to trust your own ears, not others (neither mine nor those self-proclaimed "golden guru audiophile" ones).My advice is not to trust your own ears (which easily introduce placebo), rather do your own blind listening testings (which eliminates "trust" and gives actual results), or check blind group listening test results..

In the latest group 128kbps AAC cbr blind test, FAAC was the only one which actually used ABR, instead of CBR like the other codecs, still FAAC was clearly the worst. FAAC 1.18 is no big jump in quality.