Skip to main content

Topic: New Public Listening Test (July-August 2011) (Read 50159 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • /mnt
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #50
For the AAC test i would chose:

Nero AAC
QuickTime AAC / True VBR
FhG AAC
DivX labs AAC encoder or Winamp's AAC encoder (pre-test)
FAAC

For a multi codec test, i would chose:

Winner of the AAC test.
CELT
Vorbis AuTov
LAME 3.99 beta with --vbr-new
Musepack
"I never thought I'd see this much candy in one mission!"

  • romor
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #51
for multicodec:

lame
aoTuV
AAC
WMA

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #52
I am not working for Nero anymore.

oh. I wish You the best in your new activities.


Please, check your votes and tell me if there is any error.
Now the numbers of suggestions are acceptable.
Current votes:

Multiformat - Zarggg, Kwanbis, dsimcha, Alexxander, romor (5)
AAC - Kennedyb4, /mnt, muaddib, Steve Forte Rio, IgorC, Pri3st, johnb, Polar, LithosZA, richard123 - (10)
Multiformat / but realizes that it will be difficult to choose best encoder per format - Corone (1)

Not sure what Kohlrabi wants.

Codecs:

MP3 LAME – Kwanbis, Nick.C, LithosZA, IgorC, Alexxander, richard123, /mnt, romor  (8)
Vorbis AoTuV – Kwanbis, LithosZA, IgorC, Steve Forte Rio, Alexxander (opt.), /mnt, romor (6 +1 optionally)
CELT – Steve Forte Rio, LithosZA, IgorC, /mnt (4)
WMA/WMA Pro – Kwanbis, Steve Forte Rio, romor (3)
Musepack (MPC) - /mnt (1)

Nero – Zarggg, Steve Forte Rio, IgorC, Polar, Alexxander, richard123, /mnt (7)
iTunes AAC (constrained VBR) - Kennedyb4, Zarggg, Kwanbis, IgorC, Polar, Alexxander (optinally) (5 + 1 optionally )
QuickTime AAC (true VBR) – Kennedyb4, Steve Forte Rio, Alexxander, richard123, /mnt (5)
FhG AAC - /mnt, Steve Forte Rio, IgorC, richard123 (4)
Winamp’s CT AAC – Steve Forte Rio, IgorC, /mnt (3)
FAAC – Kohlrabi, /mnt (1)
Divx AAC - /mnt (1)

Old encoder with higher bitrate – Notat, Corone, Polar, (-)lvqcl (3 + 1 against )
To include only one Apple encoder in AAC test (CVBR or TVBR) – Benski (1 developer)

Who else?

Edit: there was one error in vote list.
  • Last Edit: 26 June, 2011, 02:33:02 PM by IgorC

  • muaddib
  • [*][*][*][*]
  • Developer
New Public Listening Test (July-August 2011)
Reply #53
IMO it is a waste of time to test FAAC and DivX, and it would be better to have only one iTunes.
Maybe a pretest for iTunes or a suggestion from an Apple developer?

So for me, it is: Nero, FhG, CT and iTunes (either CVBT or TVBR).

  • lvqcl
  • [*][*][*][*][*]
  • Developer
New Public Listening Test (July-August 2011)
Reply #54
Vote for AAC test.

And I also have a question: what software uses/will use new FhG encoder?

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #55
@muaddib and lvqcl

I will update the list of votes with your choices later.

IMO it is a waste of time to test FAAC and DivX, and it would be better to have only one iTunes.
Maybe a pretest for iTunes or a suggestion from an Apple developer?

People show a little interest in FAAC and Divx. It will be safe to just exclude them.
Now it makes me think that pre-test is a bad idea. If one codec will pass the pre-test then later listeners will be more familiarized with its specific artifacts. So test will be biased against this particular codec. This has happened in one of Ivan's public test where  high anchor (MP3) was tested twice. It had no any impact as it was only high anchor. http://www.hydrogenaudio.org/forums/index....c=41191&hl=

BTW I've seen your work on perception of sound  . Interesting.

Vote for AAC test.

And I also have a question: what software uses/will use new FhG encoder?

I can't comment right now as it's not my competence. Unless developers will do their own statements. 
Though as I already have mentioned it will be available for all users on 30th of June.


I will ask people to not discuss the quality of any of the codecs right now to avoid any kind of personal prejudices during the ABC/HR sessions of this public test.

P.S. There is a lot of interest to see CVBR (popular because of  itunes) against TVBR (possible superiority).
  • Last Edit: 27 June, 2011, 05:44:16 AM by IgorC

  • muaddib
  • [*][*][*][*]
  • Developer
New Public Listening Test (July-August 2011)
Reply #56
People that have interest can also do a listening test comparing iTunes CVBR and TVBR after the first AAC listening test.
Then also another AAC encoder (or the same high anchor if there will be one) could be included in that CVBR vs TVBR test to get a scaling comparable to the first AAC test.

@IgorC: Thanks to all your nice words

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #57
@muaddib
It's ok  for me to include only one (CVBR or TVBR) mode. Or both of them.
Let's see the suggestions of other members.
Anyway we will return to this question next week during bitrate verification. TVBR and CVBR produce different bitrates.

Current votes
http://listening-test.blogspot.com/2011/06...reparation.html
If member hasn't specify whether he suggests CVBR or TVBR then it's half vote for each codec (0.5 vote for CVBR, 0.5 vote for TVBR)

People, what do you think about CVBR and TVBR? To include both of them or just one? If only one then which (CVBR or TVBR)?
But we can test only 4 codecs.
Suggestions, please.

Basic plan:
Decide what to test (AAC or Multiformat) and have an idea what codecs - until July 1
Bitrate verification  - 7-10 days - until July 7-10
Sample selection - 7-10 days  - until July 15-20
Checking all conditions, preparations, dummy packages etc.  Developers can also take a look if everything is alright. - until July 23-24
Start of test - last week of July.
  • Last Edit: 28 June, 2011, 03:40:31 AM by IgorC

New Public Listening Test (July-August 2011)
Reply #58
Quote
People, what do you think about CVBR and TVBR? To include both of them or just one? If only one then which (CVBR or TVBR)?


Only one, of course. I suppose, it should be TVBR

  • Antonski
  • [*][*][*]
New Public Listening Test (July-August 2011)
Reply #59
For a multi codec test, i would chose:

Winner of the AAC test.
CELT
Vorbis AuTov
LAME 3.99 beta with --vbr-new
Musepack


At the target bitrate of 96 kbps Musepack would be clearly inferior, I believe. It is just not supposed to be transparent there.
-1 vote

  • menno
  • [*][*][*][*][*]
  • Developer (Donating)
New Public Listening Test (July-August 2011)
Reply #60
I don't think it's a waste of time to test FAAC, by far most AAC content is encoded with it, especially in the 96kbps range. Testing FAAC can provide clear and up to date argumentation for why many internet services should or shouldn't keep using a free solution as opposed to a commercial solution.
Of course my vote doesn't count, but just trying to provide some arguments for other people to make a decision

  • /mnt
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #61
Just been forced to use ffmpeg's 0.8 aac encoder (libav52.xxx) while i was transcoding some videos for the iPod, it's a perfect canidate for a low anchor AAC encoder.

For a multi codec test, i would chose:

Winner of the AAC test.
CELT
Vorbis AuTov
LAME 3.99 beta with --vbr-new
Musepack


At the target bitrate of 96 kbps Musepack would be clearly inferior, I believe. It is just not supposed to be transparent there.
-1 vote

So is MP3.
  • Last Edit: 28 June, 2011, 02:39:21 PM by /mnt
"I never thought I'd see this much candy in one mission!"

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #62
Just been forced to use ffmpeg's 0.8 aac encoder (libav52.xxx) while i was transcoding some videos for the iPod,

MeGui has a lot of profiles (iPod as well) and uses high quality encoders x264 and Nero AAC.


I don't think it's a waste of time to test FAAC, by far most AAC content is encoded with it, especially in the 96kbps range. Testing FAAC can provide clear and up to date argumentation for why many internet services should or shouldn't keep using a free solution as opposed to a commercial solution.
Of course my vote doesn't count, but just trying to provide some arguments for other people to make a decision


it's a perfect canidate for a low anchor AAC encoder.


FAAC could be low anchor maybe at 65-70 kbps (?). Low anchor should have clearly inferior quality. It's important at the time to analyze the results.

Current votes:
http://listening-test.blogspot.com/2011/06...reparation.html
  • Last Edit: 28 June, 2011, 03:49:38 PM by IgorC

  • /mnt
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #63
Just been forced to use ffmpeg's 0.8 aac encoder (libav52.xxx) while i was transcoding some videos for the iPod,

MeGui has a lot of profiles (iPod as well) and uses high quality encoders x264 and Nero AAC.



I have tried MeGui, but i prefer to just using CLI encoders for video, such as ffmpeg and menconder. I have been working on a FFmpeg frontend in C#, which is designed to have a basic load file and select preset interface, however the Windows builds of ffmpeg on the internet are usually out of date or lacking libs such as libfaac or libx264.
"I never thought I'd see this much candy in one mission!"

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #64
ffmpeg has its own AAC encoder which is not FAAC. Last time I checked these two encoders have comparable quality, don't they?
  • Last Edit: 28 June, 2011, 04:24:20 PM by IgorC

  • /mnt
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #65
ffmpeg has its own AAC encoder which is not FAAC. Last time I checked these two encoders have comparable quality, don't they?

ffmpeg's AAC encoder sounds far worse then FAAC to me:

Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.1.7
2011/06/29 00:41:42

File A: C:\Temp\Derezzed_ffmpegAAC.m4a
File B: C:\Temp\Derezzed_FAAC.m4a

00:41:42 : Test started.
00:41:56 : 01/01  50.0%
00:42:03 : 02/02  25.0%
00:42:14 : 03/03  12.5%
00:42:22 : 04/04  6.3%
00:42:31 : 05/05  3.1%
00:42:40 : 06/06  1.6%
00:42:46 : 07/07  0.8%
00:42:50 : 08/08  0.4%
00:42:59 : 09/09  0.2%
00:43:07 : 10/10  0.1%
00:43:13 : 11/11  0.0%
00:43:19 : 12/12  0.0%
00:43:26 : 13/13  0.0%
00:43:33 : 14/14  0.0%
00:43:39 : 15/15  0.0%
00:43:47 : 16/16  0.0%
00:43:54 : 17/17  0.0%
00:43:56 : Test finished.

 ----------
Total: 17/17 (0.0%)

Both tracks are encoded at 128kbps ABR.

The ffmpeg encode has very harsh disortion and warbling. File B sounds better, but the warbling and the pre-echo at the start of the track makes it far from transparent.
"I never thought I'd see this much candy in one mission!"

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #66
Thank you, /mnt

What build do you use and settings?
Low anchor probably should be AAC with low quality and bitrate 96 or less.

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #67

  • Kef
  • [*][*][*]
New Public Listening Test (July-August 2011)
Reply #68
Multiformat gets my vote, I would be intrested to see aac, vorbis, lame and celt. But then again, why not get the best of both worlds and test a few aac encoders at the same time?

  • Kohlrabi
  • [*][*][*][*][*]
  • Global Moderator
New Public Listening Test (July-August 2011)
Reply #69
I'm with Kef here, I vote for a multiformat test with multiple AAC encoders. Something like LAME, AoTuV, CELT, iTunes AAC, Nero AAC. (was there a listening test with Musepack SV8 in it, yet?)
It's only audiophile if it's inconvenient.

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #70
@Kef and Kohlrabi
Thank you for you votes. I will update the list of votes today.


Why not get the best of both worlds and test a few aac encoders at the same time?

This is what we are trying to figure out. The votes speak for themselves.  A lot of people want AAC test to figure out more/most optimal AAC encoder(s). At the same time  there are enough people who want to see multiformat test.
What is your suggestion for selection of AAC encoders and what are more important factors for such decisions (popularity, quality and availability of encoder)?


Something like LAME, AoTuV, CELT, iTunes AAC, Nero AAC. (was there a listening test with Musepack SV8 in it, yet?)

5-6 codecs are too much for one test. Everything will be tied.
We will hardly get statistical difference with such high amount of codecs.

4 is still affordable.


@/mnt
If ffmpeg's AAC encoder is pretty bad then 128 kbps will be good.
What do you think?
Then it will fulfill this request. Two birds with one stone.
Quote
Old/low quality encoder with higher bitrate – Notat, Corone, Polar, (-)lvqcl (3 + 1 against )
  • Last Edit: 29 June, 2011, 07:47:17 AM by IgorC

  • /mnt
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #71
Thank you, /mnt

What build do you use and settings?
Low anchor probably should be AAC with low quality and bitrate 96 or less.

I was a using Windows build from git at 26/06/2009, which lacked libfaac support.

The settings was:

ffmpeg -i file -acodec aac -ab 128k -strict experimental file.m4a

@/mnt
If ffmpeg's AAC encoder is pretty bad then 128 kbps will be good.
What do you think?
Then it will fulfill this request. Two birds with one stone.

ffmpeg's AAC encoder seems to produce far worse results then FAAC, it's AAC's answer to Blade.

I have done a new test with the offical stable release of ffmpeg 0.8 on Ubuntu 10.04 (my own compile with libx264, libxv264, libfaac and libmp3lame enabled).

ffmpeg AAC settings:

ffmpeg -i file -acodec aac -ab 128k -strict experimental file.m4a

ffmpeg FAAC settings:

ffmpeg -i file -acodec libfaac -ab 128k file-faac.m4a


Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.1.7
2011/06/29 15:42:18

File A: C:\Temp\Linchpin (ffmpeg AAC).m4a
File B: C:\Temp\Linchpin (ffmpeg FAAC).m4a

15:42:18 : Test started.
15:42:30 : 01/01  50.0%
15:42:35 : 02/02  25.0%
15:42:41 : 03/03  12.5%
15:42:48 : 04/04  6.3%
15:42:53 : 05/05  3.1%
15:42:57 : 06/06  1.6%
15:43:05 : 07/07  0.8%
15:43:08 : 08/08  0.4%
15:43:13 : 09/09  0.2%
15:43:16 : 10/10  0.1%
15:43:21 : 11/11  0.0%
15:43:31 : 12/12  0.0%
15:43:32 : Test finished.

 ----------
Total: 12/12 (0.0%)

Warbling and disortion on the aac encode, while faac has less harsh pre-echo but some noticable tonal disortion. However faac sounds far better.

Code: [Select]
foo_abx 1.3.4 report
foobar2000 v1.1.7
2011/06/29 15:44:06

File A: C:\Temp\Through Being Cool (ffmpeg AAC).m4a
File B: C:\Temp\Through Being Cool (ffmpeg FAAC).m4a

15:44:06 : Test started.
15:44:19 : 01/01  50.0%
15:44:24 : 02/02  25.0%
15:44:28 : 03/03  12.5%
15:44:33 : 04/04  6.3%
15:44:43 : 05/05  3.1%
15:44:49 : 06/06  1.6%
15:44:55 : 07/07  0.8%
15:44:59 : 08/08  0.4%
15:45:05 : 09/09  0.2%
15:45:13 : 10/10  0.1%
15:45:18 : 11/11  0.0%
15:45:21 : 12/12  0.0%
15:45:22 : Test finished.

 ----------
Total: 12/12 (0.0%)

The aac encode has horrid warbling all over the synth at the start, also it produces sharp disortion and drop outs when the lead vocals pauses. FAAC sounds better, however the vocals can produce disortion and also it struggles to cope with the recording artifacts that appear on the track.
  • Last Edit: 29 June, 2011, 11:06:03 AM by /mnt
"I never thought I'd see this much candy in one mission!"

  • IgorC
  • [*][*][*][*][*]
New Public Listening Test (July-August 2011)
Reply #72
Yep, I have tried for myself. ffmpeg AAC produces very inferior quality even at 128 kbps.  It will be our low anchor. 

FAAC is far superior. In fact FAAC isn't that bad at all. But there are a lot of good (commercial) AAC encoders for one test.

Current votes will be updated shortly:
http://listening-test.blogspot.com/2011/06...reparation.html
  • Last Edit: 29 June, 2011, 03:41:45 PM by IgorC

  • benski
  • [*][*][*][*][*]
  • Developer
New Public Listening Test (July-August 2011)
Reply #73
Vote for AAC test.

And I also have a question: what software uses/will use new FhG encoder?


Winamp, starting with v5.62 (released today)

  • lvqcl
  • [*][*][*][*][*]
  • Developer
New Public Listening Test (July-August 2011)
Reply #74
Wow. Downloading right now.