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: Snow Leopard AAC improvements.. (Read 73335 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Snow Leopard AAC improvements..

Reply #50
Regarding the ipod + HE-AAC discussion.. What I would like to know is whether those of us who have their library stored in lossless format and convert the songs to a lossy format only for portable players really need higher quality than HE-AAC has to offer.


HE-AAC does not target transparency! It just targets to be the better sounding "good enough" encoder for very low bitrates. For transparency >=128 kbit/s LC-AAC is still the much better choice.



Yes that's exactly what I said. But do you need transparency in a noisy environment? Do you think it will be possible to hear the difference? And if so what would be your choice regarding AAC bitrate for an ipod? (while having your library stored in FLAC).

Snow Leopard AAC improvements..

Reply #51
Regarding the ipod + HE-AAC discussion.. What I would like to know is whether those of us who have their library stored in lossless format and convert the songs to a lossy format only for portable players really need higher quality than HE-AAC has to offer.


HE-AAC does not target transparency! It just targets to be the better sounding "good enough" encoder for very low bitrates. For transparency >=128 kbit/s LC-AAC is still the much better choice.


For the average Joe listening to Music on the go in a noisy place like a train or on the street, HE-AAC provides the right balance. I wouldn't be surprised if one ended up seeing wider usage since most people's hearing is shot to pieces or are unconcerned about music quality over all.

Snow Leopard AAC improvements..

Reply #52
So what is the verdict on the new AAC encoder? Judging by XLD's new True VBR default, I gather that q90 is supposed to take the place of q127 as far as file sizes that are produced. But are the tweaks to the encoder such that it is actually producing higher-quality tracks than the old q127?
Altec UHP336s/Sennheiser eH 150s/AKG K171s & q113 tVBR AAC...

Snow Leopard AAC improvements..

Reply #53
Interesting - I can easily ABX emese at Q127:


Mixminus1 (or anyone else with extreme abx gifts, skills),
were you able to abx that sample or any sample with Nero at any setting (including the highest, q 1.0)?   

I wanted to confirm that it's still true that no aac or mp3 encoder can yet pass every killer sample regardless of setting (I mean at any setting less than the bitrate lossless would produce).  I've seen this confirmed with lame and (by mixminus1) with the recent quicktime aac release, but not yet with the latest nero.  I'd try myself, but I have mediocre abx skills, so I lack the power.  Thank you.

Snow Leopard AAC improvements..

Reply #54
Judging from the Windows version of Quick Time's True VBR via qtaacenc.exe, i find it to be quite disapointing with tracks that produce heavy pre-echo artifacts. I find --tvbr 100 to perform quite poorly with a few Kraftwerk tracks i tested, and --tvbr 127 dosn't make much difference.

A couple of examples:

Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/16 20:02:25

File A: J:\Music\Lossless\Kraftwerk - Electric Café\03. Musique Non Stop.flac
File B: C:\Temp\Musique Non Stop QT TVBR 127.m4a

20:02:25 : Test started.
20:02:44 : 01/01  50.0%
20:02:48 : 02/02  25.0%
20:02:52 : 03/03  12.5%
20:02:57 : 04/04  6.3%
20:03:01 : 05/05  3.1%
20:03:06 : 06/06  1.6%
20:03:10 : 07/07  0.8%
20:03:16 : 08/08  0.4%
20:03:21 : 09/09  0.2%
20:03:28 : 10/10  0.1%
20:03:32 : 11/11  0.0%
20:03:36 : 12/12  0.0%
20:03:41 : 13/13  0.0%
20:03:45 : 14/14  0.0%
20:03:50 : 15/15  0.0%
20:03:56 : 16/16  0.0%
20:04:01 : 17/17  0.0%
20:04:04 : 18/18  0.0%
20:04:11 : 19/19  0.0%
20:04:16 : 20/20  0.0%
20:04:22 : 21/21  0.0%
20:04:25 : 22/22  0.0%
20:04:31 : 23/23  0.0%
20:04:35 : 24/24  0.0%
20:04:41 : 25/25  0.0%
20:04:47 : 26/26  0.0%
20:04:50 : 27/27  0.0%
20:04:56 : 28/28  0.0%
20:04:59 : 29/29  0.0%
20:05:04 : 30/30  0.0%
20:05:06 : Test finished.

 ----------
Total: 30/30 (0.0%)

Loud disortion at 0:18.


Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/16 20:10:12

File A: J:\Music\Lossless\Kraftwerk - The Man Machine\01. The Robots.flac
File B: C:\Temp\The Robots QT TVBR 127.m4a

20:10:12 : Test started.
20:10:48 : 01/01  50.0%
20:10:59 : 02/02  25.0%
20:11:08 : 03/03  12.5%
20:11:14 : 04/04  6.3%
20:11:30 : 05/05  3.1%
20:11:38 : 06/06  1.6%
20:11:48 : 07/07  0.8%
20:12:07 : 08/08  0.4%
20:12:17 : 09/09  0.2%
20:12:28 : 10/10  0.1%
20:12:40 : 11/11  0.0%
20:12:52 : 12/12  0.0%
20:12:54 : Test finished.

 ----------
Total: 12/12 (0.0%)

Disortion at 0:31 - 0:32.

Those tracks were encoded by QuickTime 7.6.5 via foobar2000 1.01 + qtaacenc.exe at --tvbr 127 --highest.
"I never thought I'd see this much candy in one mission!"

Snow Leopard AAC improvements..

Reply #55
Judging from the Windows version of Quick Time's True VBR via qtaacenc.exe, i find it to be quite disapointing with tracks that produce heavy pre-echo artifacts. I find --tvbr 100 to perform quite poorly with a few Kraftwerk tracks i tested, and --tvbr 127 dosn't make much difference.
Thank you /mnt, it's good to have a second to confirm that and giving the exact breakdown points is very helpful.  I had assumed pre-echo was one thing aac was supposed to be much better than mp3 at, and especially at such super high rates as ~320kbps vbr.  (otoh kraftwerk is synth among synths.) 

I don't know if you've had or have time to do this, but have you (or anyone else) been able to abx with the new Nero aac encoder at the most generous settings?  Does even the highest setting, q1.00 (I think ~380-420 kbps), fail for you (and as poorly)?  (And if no lossy encoder can escape detection regardless of how high the bitrate is set, short of lossless bitrates, perhaps it would now escape detection with non-synthetic music? or maybe I have to be wary of lossy encodes of my harpsichord-castanet concerts.)

Snow Leopard AAC improvements..

Reply #56
Does even the highest setting, q1.00 (I think ~380-420 kbps), fail for you (and as poorly)?  (And if no lossy encoder can escape detection regardless of how high the bitrate is set, short of lossless bitrates, perhaps it would now escape detection with non-synthetic music? or maybe I have to be wary of lossy encodes of my harpsichord-castanet concerts.)

I cannot ABX those samples with Nero AAC at q 1.0, but there is a track from Electric Cafe / Techno Pop that Nero AAC struggles with, even at q 1.0.


Nero AAC -q 0.50 (1.5.3.0)
Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/20 14:18:43

File A: C:\Music\Kraftwerk\Electric Café\02. Techno Pop.m4a
File B: J:\Music\Lossless\Kraftwerk - Electric Café\02. Techno Pop.flac

14:18:43 : Test started.
14:20:07 : 01/01  50.0%
14:20:17 : 02/02  25.0%
14:20:26 : 03/03  12.5%
14:20:33 : 04/04  6.3%
14:20:41 : 05/05  3.1%
14:20:48 : 06/06  1.6%
14:20:56 : 07/07  0.8%
14:21:08 : 08/08  0.4%
14:21:26 : 09/09  0.2%
14:21:35 : 10/10  0.1%
14:21:51 : Test finished.

 ----------
Total: 10/10 (0.1%)
Pre-echo at 3.57 and 4:01.

q 0.80 (Nero AAC 1.5.4.0)
Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/20 14:30:39

File A: J:\Music\Lossless\Kraftwerk - Electric Café\02. Techno Pop.flac
File B: C:\Temp\Techno Pop Nero AAC q q.80.m4a

14:30:39 : Test started.
14:31:09 : 01/01  50.0%
14:31:22 : 02/02  25.0%
14:31:35 : 03/03  12.5%
14:31:57 : 04/04  6.3%
14:32:19 : 05/05  3.1%
14:32:31 : 06/06  1.6%
14:32:39 : 07/07  0.8%
14:32:54 : 08/08  0.4%
14:33:14 : 09/09  0.2%
14:33:35 : 10/10  0.1%
14:33:45 : 11/11  0.0%
14:33:58 : 12/12  0.0%
14:34:05 : Test finished.

 ----------
Total: 12/12 (0.0%)
The artifacts are softer, which required more concerntration. It's still far from transparent though.

q 1.0 (Nero AAC 1.5.4.0)
Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/20 14:35:43

File A: J:\Music\Lossless\Kraftwerk - Electric Café\02. Techno Pop.flac
File B: C:\Temp\Techno Pop Nero AAC q 1.0.m4a

14:35:43 : Test started.
14:36:52 : 01/01  50.0%
14:37:10 : 02/02  25.0%
14:37:23 : 03/03  12.5%
14:37:37 : 04/04  6.3%
14:38:02 : 05/05  3.1%
14:38:16 : 06/06  1.6%
14:38:53 : 07/07  0.8%
14:39:04 : 08/08  0.4%
14:39:16 : 09/09  0.2%
14:39:40 : 10/10  0.1%
14:40:21 : 11/11  0.0%
14:40:38 : 12/12  0.0%
14:40:53 : Test finished.

 ----------
Total: 12/12 (0.0%)
Artifact at 3:57 is much more softer and almost disappears, but the artifact 4:01 is still obivous.

Also i have tried out QuickTime AAC at --tvbr 127:

Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.0.1
2010/03/20 15:10:51

File A: J:\Music\Lossless\Kraftwerk - Electric Café\02. Techno Pop.flac
File B: C:\Temp\Techno Pop QT TVBR 127.m4a

15:10:51 : Test started.
15:11:12 : 01/01  50.0%
15:11:30 : 02/02  25.0%
15:11:45 : 03/03  12.5%
15:11:53 : 04/04  6.3%
15:12:12 : 05/05  3.1%
15:12:17 : 06/06  1.6%
15:12:33 : 07/07  0.8%
15:13:12 : 08/08  0.4%
15:13:18 : 09/09  0.2%
15:13:28 : 10/10  0.1%
15:13:34 : Test finished.

 ----------
Total: 10/10 (0.1%)
The synth at 3:57 sounds more sharp on the QuickTime encode, but this was more harder to ABX then Nero AAC at -q 80.
"I never thought I'd see this much candy in one mission!"

Snow Leopard AAC improvements..

Reply #57
The synth at 3:57 sounds more sharp on the QuickTime encode, but this was harder to ABX then Nero AAC at -q 80.
At nero -q 0.80--but I assume that at -q 1.0 you so far find (which would not be surprising given the higher bitrate, 400+ kbps), that nero is always equal or better quality than quicktime at its max (which at -tvbr 127 tops around 320kbps). 
(re  Techno Pop with nero -q 1.0) Artifact at 3:57 is much more soft and almost disappears, but the artifact at 4:01 is still obvious.
wow.  obvious artifacts to someone even at 400+ kbps aac!  well, I'm still waiting to see if someone can find obvious artifacts at nero -q 1.0 or even quicktime -tvbr 127 with non-synthetic music (maybe harpsichords?). 

Thank you /mnt for posting your abx test results--even with such fine hearing, I know it's more time-consuming and much harder to do these tests as the bitrate goes up so high.  (It's good also that you confirm that higher bitrates, even at already superhigh rates, do make a difference in softening if not always eliminating the artifact--perhaps it's increasingly smaller reward for the higher bitrates, but still some reward.)

p.s. is there a link to download the technopop sample?  I didn't find it.


Snow Leopard AAC improvements..

Reply #59
Thanks, got it.  (and it does sound robotic.)

Snow Leopard AAC improvements..

Reply #60
Shouldn't this be submitted as a bug report to Apple?

Snow Leopard AAC improvements..

Reply #61
Shouldn't this be submitted as a bug report to Apple?

It's not exactly a bug, though should be reported to apple.  It's just an issue of whether the aac encoder could be tuned better to soften or eliminate those artifacts (without creating another problem elsewhere).  I assume that there still is significant room for improvement in aac encoding, but I don't know if it matters to their customers much.  Also remember these are synthetic/electronic music pieces and especially known to be "killers" (not representative of the vast majority of music pieces, which probably would not be abx'able at max. aac bitrates encoded by current apple/nero aac encoders -- *at least, I found no reports yet*).  (imo one has to be a masochist to enjoy listening to "emese" sample.)
fyi there's apple lossless format, guaranteed 100% faithful, but for some it's a hassle to maintain and manage two formats for everything--and lossless is still too big given current storage capacities on portable players and their battery life.  also no lossless format is widely and comfortably supported by many types of hardware and software anyway--mp3 and aac are.

Snow Leopard AAC improvements..

Reply #62
Perhaps soon a new QuickTime for Windows? 

Quote
Apple has provided developers with a second private beta of Mac OS X 10.6.3, hinting that an update to Snow Leopard will be available to the public soon.

The latest update to Snow Leopard to be given to developers has the build number of 10D572; this comes just days after Apple gave developers a pre-release with the build number 10D571.

According to people familiar with the beta versions, Apple is now asking developers to focus on certain applications including: Mail, images, QuickTime and graphic drivers. In the 10D571 beta version of Snow Leopard, developers were focusing on fonts and iCal.

When Mac OS X 10.6.3 is finally released there will also be an update for QuickTime that enables improved security and compatibility that will improve the overall performance of the software.

Snow Leopard AAC improvements..

Reply #63
There's been a while since the last major update to the AAC encoder so it should be interesting to test this new version when it hits the streets.

Snow Leopard AAC improvements..

Reply #64
So Mac OS X 10.6.3 with Quicktime 7.6.6 is out, and the AAC encoder is definitely updated as the bitrates are different from the previous version.. Should be interesting to see how it compares on the killer samples.

Maybe Windows users will get QT 7.6.6 with iTunes 9.1 rumored to debut soon?

Snow Leopard AAC improvements..

Reply #65
Is it just me or the new 7.6.6 has improved even further quality? Before there were some songs i could easily abx but with this it's a lot harder to abx,at least for me. Great job Apple,really great!

I'm pretty sure Windows and leopard users will get this with iTunes 9.1 in a few days.

Snow Leopard AAC improvements..

Reply #66
iTunes 9.1 might be released with the iPad.  To my knowledge, the current version does not support the iPad so Apple would have to roll out an updated version.  I don't know if QuickTime 7.6.6 would be part of that or even if iTunes needs to be updated for the iPad.  As I said, I have not heard anything regarding the current edition dealing with iPad compatibility.

Snow Leopard AAC improvements..

Reply #67
So Mac OS X 10.6.3 with Quicktime 7.6.6 is out, and the AAC encoder is definitely updated as the bitrates are different from the previous version..


I just downloaded QT 7.6.6 (updating from 7.6.5) and the bitrates were exactly the same! However, this was on Windows (using nao's qtaacenc) rather than Mac....

Snow Leopard AAC improvements..

Reply #68
Hmm I'm seeing bitrate changes on the tracks I've checked, although sometimes the variations are so slight that the files have to be compared in QT which shows the bitrates with decimals..

In any case, md5 checksums differ from the previous files after decoding them to wav, so there's definitely been a update from QT 7.6.3 (the most recent on Snow Leopard).

Snow Leopard AAC improvements..

Reply #69
I installed the new quicktime 7.6.6 on windows.  Using qtaacenc.exe and the max quality setting (-highest -tvbr 127), I show no binary difference in the resulting music data encoded (though the "encoded by" metadata changed from 7.6.5 to 7.6.6).

Snow Leopard AAC improvements..

Reply #70
i used XLD on Mac and it showed different bitrates instead,i don't know.

Snow Leopard AAC improvements..

Reply #71
I can confirm this on Windows, using nao's qtaacenc, file size for true vbr quality 100 is now about 5% bigger, which probably has improved the quality somewhat.

Snow Leopard AAC improvements..

Reply #72
The changelogs, at least those available to the public, show no mention of the AAC encoder.

 

Snow Leopard AAC improvements..

Reply #73
FYI, an internal codec version is
  • 7.6.3: 1.7.0(LC), 1.0.0(HE)
  • 7.6.4: 1.7.1(LC), 1.2.0(HE)
  • 7.6.5: 1.7.1(LC), 1.2.0(HE)
  • 7.6.6: 1.7.2(LC), 1.2.1(HE)

Snow Leopard AAC improvements..

Reply #74
I can confirm this on Windows, using nao's qtaacenc, file size for true vbr quality 100 is now about 5% bigger, which probably has improved the quality somewhat.

What version are you comparing it to? (what version of Quicktime were you using before you upgraded?)  I don't see any confirmed change in the music data produced from the encoder when upgrading from 7.6.5 to 7.6.6.  Note also I'm on Windows.