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: Possible LAME 3.97 -V 2 killer samples. (Read 21040 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Possible LAME 3.97 -V 2 killer samples.

Reply #25
I found a artifact on another mp3 i have on my computer that has been encoded on LAME 3.97 -V 2 --vbr-new.

LAME 3.97 -V 2 --vbr-new

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/09 22:30:24

File A: F:\Listen Tests\Enter Sandman.flac
File B: E:\Music\Albums\Metallica - Metallica\01. Metallica - Enter Sandman.mp3

22:30:24 : Test started.
22:31:51 : 01/01  50.0%
22:32:09 : 02/02  25.0%
22:32:25 : 03/03  12.5%
22:32:52 : 04/04  6.3%
22:33:19 : 05/05  3.1%
22:33:38 : 06/06  1.6%
22:33:49 : 07/07  0.8%
22:34:25 : 08/08  0.4%
22:34:44 : 09/09  0.2%
22:34:55 : 10/10  0.1%
22:35:28 : 11/11  0.0%
22:35:45 : 12/12  0.0%
22:36:00 : Test finished.

----------
Total: 12/12 (0.0%)


A precho at drums at around 4:22.

LAME 3.98 Beta 5 -V 2

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/09 22:59:40

File A: F:\Listen Tests\Enter Sandman.flac
File B: F:\Listen Tests\Enter Sandman V2 LAME3.98b5.mp3

22:59:40 : Test started.
23:00:54 : 00/01  100.0%
23:01:21 : 01/02  75.0%
23:01:36 : 01/03  87.5%
23:02:42 : 02/04  68.8%
23:03:04 : 02/05  81.3%
23:03:48 : 03/06  65.6%
23:05:09 : 04/07  50.0%
23:05:21 : 05/08  36.3%
23:06:16 : 06/09  25.4%
23:06:38 : 07/10  17.2%
23:07:27 : 08/11  11.3%
23:08:00 : 08/12  19.4%
23:08:32 : Test finished.

----------
Total: 8/12 (19.4%)


LAME 3.90.3 --alt-preset standard

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/09 22:51:41

File A: F:\Listen Tests\Enter Sandman.flac
File B: F:\Listen Tests\Enter Sandman LAME3.90 APS.mp3

22:51:41 : Test started.
22:52:55 : 01/01  50.0%
22:53:39 : 01/02  75.0%
22:54:07 : 02/03  50.0%
22:54:27 : 02/04  68.8%
22:54:35 : 03/05  50.0%
22:54:51 : 03/06  65.6%
22:55:07 : 04/07  50.0%
22:55:22 : 05/08  36.3%
22:55:49 : 06/09  25.4%
22:56:06 : 06/10  37.7%
22:56:13 : 07/11  27.4%
22:56:34 : 08/12  19.4%
22:57:20 : Test finished.

----------
Total: 8/12 (19.4%)
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #26
hey /mnt... can you ABX these with Nero AAC at -q 5.5

Possible LAME 3.97 -V 2 killer samples.

Reply #27
hey /mnt... can you ABX these with Nero AAC at -q 5.5

I'll give it a try.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #28
I did a quick test with Nero AAC encoder.

I used:

Nero AAC codec / Feb 12 2007

since I found a killer sample on the latest version.

http://www.hydrogenaudio.org/forums/index....showtopic=56849

Nero AAC codec / Feb 12 2007 -q0.55

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/09 23:34:06

File A: F:\Listen Tests\Until It Sleeps.wav
File B: F:\Listen Tests\Until It Sleeps.mp4

23:34:06 : Test started.
3:35:22 : 00/01  100.0%
23:35:48 : 00/02  100.0%
23:36:02 : 01/03  87.5%
23:36:31 : 02/04  68.8%
23:36:44 : 03/05  50.0%
23:37:01 : 03/06  65.6%
23:37:34 : 04/07  50.0%
23:38:06 : 04/08  63.7%
23:38:35 : 04/09  74.6%
23:38:48 : 04/10  82.8%
23:39:01 : 05/11  72.6%
23:39:14 : 05/12  80.6%
23:39:17 : Test finished.

----------
Total: 5/12 (80.6%)


I had trouble with that one.

Nero AAC codec / Feb 12 2007 -q0.55

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/09 23:49:53

File A: F:\Listen Tests\Enter Sandman.flac
File B: F:\Listen Tests\Enter Sandman.mp4

23:49:53 : Test started.
23:51:00 : 01/01  50.0%
23:51:35 : 01/02  75.0%
23:51:52 : 02/03  50.0%
23:52:30 : 03/04  31.3%
23:52:45 : 03/05  50.0%
23:52:54 : 04/06  34.4%
23:53:28 : 05/07  22.7%
23:54:06 : 06/08  14.5%
23:54:39 : 07/09  9.0%
23:55:21 : 07/10  17.2%
23:55:49 : 08/11  11.3%
23:56:31 : 09/12  7.3%
23:56:47 : Test finished.

----------
Total: 9/12 (7.3%)


Kept hearing a weird noise at around 4:22 - 4:30 possible metallic warbling.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #29
heavy metal / hard rock is usually very difficult for me to ABX it - the genre I find the easier to ABX is synthpop/electronic... 

i feel like giving up on lossy codecs for good, and just sit here in silence on my own until -something- plays flac off an optical disc.

Possible LAME 3.97 -V 2 killer samples.

Reply #30
heavy metal / hard rock is usually very difficult for me to ABX it - the genre I find the easier to ABX is synthpop/electronic... 

i feel like giving up on lossy codecs for good, and just sit here in silence on my own until -something- plays flac off an optical disc.

Yeah Metal and Rock is hard to ABX since sometimes the drums sound like metallic warbling such as Seek And Destroy by Metallica which has warbling sound like noises like it from a 64kpbs mp3 and the high freq noises from metal tracks does not help aswell

I might move to lossless soon if a lossless format becomes a norm like mp3, which FLAC might end being a stardard lossless or Fhg MPEG4 ALS.

But at the mean time i might re-rip everything to -V0 --vbr-new or use Nero AAC at -q0.65 since it gives the bitrate results as LAME -V 0 or -q 0.6.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #31
i feel like giving up on lossy codecs for good, and just sit here in silence on my own until -something- plays flac off an optical disc.


That is a radical approach.
Does the artifact bother you that much? I mean, when listening casually artifacts are not that annoying to me (especially at V2). Most of the time I cannot hear them, only if I really concentrate.
You know, music is supposed to be a pleasant experience but if you keep listening trying to spot flaws in the encoding it becomes an unpleasant thing. Just relax and enjoy the music.

Possible LAME 3.97 -V 2 killer samples.

Reply #32
Training to spot artifacts is a one-way ticket - I'm not totally in peace until I know it's lossless

Possible LAME 3.97 -V 2 killer samples.

Reply #33
Dam I found another one, which i picked up on my crappy apple earpods from my ipod.

LAME 3.97 -V 2 --vbr-new

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/10 13:53:34

File A: F:\Listen Tests\My Friend Of Misery.flac
File B: E:\Music\Albums\Metallica - Metallica\11. Metallica - My Friend Of Misery.mp3

13:53:34 : Test started.
13:55:22 : 01/01  50.0%
13:55:52 : 02/02  25.0%
13:56:54 : 03/03  12.5%
13:57:09 : 04/04  6.3%
13:57:53 : 05/05  3.1%
13:59:02 : 06/06  1.6%
13:59:45 : 07/07  0.8%
14:00:08 : 08/08  0.4%
14:01:20 : 08/09  2.0%
14:02:32 : 09/10  1.1%
14:03:08 : 10/11  0.6%
14:03:45 : 11/12  0.3%
14:03:48 : Test finished.

----------
Total: 11/12 (0.3%)


Preecho at drums at 3:22

LAME 3.98 Beta 5 -V 2

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/10 14:12:06

File A: F:\Listen Tests\My Friend Of Misery.flac
File B: F:\Listen Tests\My Friend Of Misery LAME3.98b5 V2.mp3

14:12:06 : Test started.
14:13:51 : 01/01  50.0%
14:14:03 : 02/02  25.0%
14:14:21 : 02/03  50.0%
14:14:29 : 03/04  31.3%
14:14:59 : 04/05  18.8%
14:15:15 : 05/06  10.9%
14:15:46 : 06/07  6.3%
14:17:00 : 07/08  3.5%
14:17:25 : 07/09  9.0%
14:17:38 : 08/10  5.5%
14:18:06 : 09/11  3.3%
14:18:18 : 10/12  1.9%
14:18:24 : Test finished.

----------
Total: 10/12 (1.9%)


The preecho is still there but its harder to hear then the LAME 3.97 version.

I did a abx with the track at V0 on LAME 3.97 and got 7/12, the precho was still there but very hard to hear.

Also some of the tracks i abx showed that LAME 3.98 was better then LAME 3.97 and for some reason -V 2 is not transparent to me anymore, which is kinda odd.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #34
I wouldn't reencode to a higher mp3 setting if preecho is a major issue. Go with lossless or very high bitrate hybrid lossy 450..550 k (zero preecho). Try the new lossywav if you have time.
wavpack hybrid 256k -hx4

Possible LAME 3.97 -V 2 killer samples.

Reply #35
I have Lossless image files on DVDRs as my backup and I just copy them onto the harddrive and convert them to whatever lossy format will work best on my DAP.  Currently I use Lame 3.98b5 -v1 -vbr-new but it may change due to quality reasons (i.e. get quality with a smaller size in a different format or improvements in Lame) so I can swing with these  changes with no trouble at all by just popping in a DVD and converting the image file into whatever I need through foobar2000.

*Results may vary*

FYI,
Image file is made from a EAC Secure extraction with Cuesheet, wav file and EAC Log, if the covers are rare to find on the internet I scan and embed them into the file.


Have you tried ABXing -v1 in comparison to -v2 in 3.98b5?
Zune 80, Tak -p4 audio library, Lossless=Choice

Possible LAME 3.97 -V 2 killer samples.

Reply #36
Dam I found another one, which i picked up on my crappy apple earpods from my ipod.

How does it sound with 3.90 --alt-preset standard?

Possible LAME 3.97 -V 2 killer samples.

Reply #37
Have you tried ABXing -v1 in comparison to -v2 in 3.98b5?


I did a quick test on that setting

LAME 3.97 -V 1 --vbr-new

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/10 20:26:01

File A: F:\Listen Tests\Until It Sleeps LAME3.97 V1.mp3
File B: F:\Listen Tests\Until It Sleeps.wav

20:26:01 : Test started.
20:26:34 : 00/01  100.0%
20:26:59 : 00/02  100.0%
20:27:26 : 00/03  100.0%
20:28:10 : 01/04  93.8%
20:28:37 : 02/05  81.3%
20:29:04 : 02/06  89.1%
20:29:32 : 02/07  93.8%
20:29:46 : 02/08  96.5%
20:30:18 : 03/09  91.0%
20:30:34 : 03/10  94.5%
20:30:54 : 04/11  88.7%
20:31:44 : 04/12  92.7%
20:31:45 : Test finished.

----------
Total: 4/12 (92.7%)


The file size goes up by 1mb and a 30kbps increase. Anyway the annyoing warbling seems to have gone.

Dam I found another one, which i picked up on my crappy apple earpods from my ipod.

How does it sound with 3.90 --alt-preset standard?


LAME 3.90.3 --alt-preset standard

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/10 20:09:47

File A: F:\Listen Tests\My Friend Of Misery LAME 3.90 APS.mp3
File B: F:\Listen Tests\My Friend Of Misery.flac

20:09:47 : Test started.
20:10:48 : 01/01  50.0%
20:10:56 : 02/02  25.0%
20:11:09 : 03/03  12.5%
20:12:44 : 03/04  31.3%
20:13:08 : 03/05  50.0%
20:13:48 : 04/06  34.4%
20:13:59 : 05/07  22.7%
20:14:45 : 06/08  14.5%
20:15:17 : 07/09  9.0%
20:15:26 : 08/10  5.5%
20:16:22 : 08/11  11.3%
20:17:53 : 08/12  19.4%
20:17:56 : Test finished.

----------
Total: 8/12 (19.4%)


Its sounds good as the LAME 3.98 encode.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #38
Why is it impossible to create a 30 second extract which still demonstrates the artefact?

I understand the bit reservoir, but maybe it would be sensible to check how that is being used in encoding the full version vs encoding the 30 second clip, before assuming that's the culprit.

I'm sure one of the lame devs can point to a tool which will show what's happening.

Cheers,
David.

Possible LAME 3.97 -V 2 killer samples.

Reply #39
Another one a precho again at drums.

LAME 3.97 -V 2 --vbr-new

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/11 13:48:42

File A: F:\Listen Tests\Mr Crowley.flac
File B: E:\Music\Albums\Ozzy Osbourne - Blizzard Of Ozz [2002 Remaster]\06. Ozzy Osbourne - Mr. Crowley.mp3

13:48:42 : Test started.
13:49:53 : 01/01  50.0%
13:50:38 : 02/02  25.0%
13:51:28 : 03/03  12.5%
13:51:44 : 04/04  6.3%
13:51:53 : 05/05  3.1%
13:53:16 : 05/06  10.9%
13:54:30 : 06/07  6.3%
13:55:21 : 07/08  3.5%
13:56:38 : 08/09  2.0%
13:57:09 : 09/10  1.1%
13:57:50 : 10/11  0.6%
13:58:08 : 11/12  0.3%
13:58:34 : Test finished.

----------
Total: 11/12 (0.3%)


A precho at drums at near end of the track. My -V 2 --vbr-new transparent bubble has burst

LAME 3.98 -V 2

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/11 14:03:20

File A: F:\Listen Tests\Mr Crowley.flac
File B: F:\Listen Tests\Mr Crowley LAME 3.98 V2.mp3

14:03:20 : Test started.
14:05:43 : 01/01  50.0%
14:06:13 : 01/02  75.0%
14:06:53 : 01/03  87.5%
14:07:54 : 02/04  68.8%
14:08:17 : 02/05  81.3%
14:09:21 : 03/06  65.6%
14:09:33 : 03/07  77.3%
14:10:33 : 03/08  85.5%
14:10:41 : 03/09  91.0%
14:11:44 : 03/10  94.5%
14:11:52 : 04/11  88.7%
14:13:17 : 05/12  80.6%
14:13:20 : Test finished.

----------
Total: 5/12 (80.6%)


Lot better then LAME 3.97, I had trouble abxing that one.
"I never thought I'd see this much candy in one mission!"

Possible LAME 3.97 -V 2 killer samples.

Reply #40
Another ABX test i just did, sorry for turning this topic into a personal listen test.

The worst artifact i found so far in my mp3 collection and the artifact on that track has been bugging me for months.

LAME 3.97 -V 2 --vbr-new

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/11 22:44:02

File A: F:\Listen Tests\Stigmata.flac
File B: E:\Music\Albums\Ministry - The Land Of Rape And Honey\01. Ministry - Stigmata.mp3

22:44:02 : Test started.
22:44:43 : 01/01  50.0%
22:44:53 : 02/02  25.0%
22:45:32 : 03/03  12.5%
22:45:56 : 04/04  6.3%
22:46:19 : 05/05  3.1%
22:46:53 : 06/06  1.6%
22:47:27 : 07/07  0.8%
22:47:49 : 08/08  0.4%
22:48:25 : 09/09  0.2%
22:49:16 : 10/10  0.1%
22:49:42 : 11/11  0.0%
22:50:04 : 12/12  0.0%
22:50:48 : Test finished.

----------
Total: 12/12 (0.0%)


Worst precho artifact I heard so far on -V 2 --vbr-new, but its transparent on -V0 --vbr-new.

LAME 3.90.3 --alt-preset standard

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/11 23:21:59

File A: F:\Listen Tests\Stigmata (LAME 3.90.3 APS).mp3
File B: F:\Listen Tests\Stigmata.wav

23:21:59 : Test started.
23:23:07 : 01/01  50.0%
23:23:24 : 02/02  25.0%
23:25:09 : 02/03  50.0%
23:26:13 : 02/04  68.8%
23:26:57 : 02/05  81.3%
23:27:23 : 03/06  65.6%
23:27:33 : 04/07  50.0%
23:27:42 : 05/08  36.3%
23:27:51 : 06/09  25.4%
23:28:06 : 07/10  17.2%
23:28:54 : 08/11  11.3%
23:29:16 : 09/12  7.3%
23:29:17 : Test finished.

----------
Total: 9/12 (7.3%)


Sounds alot better then the LAME 3.97 encode but the artifact is still there but i needed to concentrate to hear  it.

LAME 3.98 Beta 5 -V 2

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/11 23:08:10

File A: F:\Listen Tests\Stigmata.wav
File B: F:\Listen Tests\Stigmata LAME 3.98b5 V2.mp3

23:08:10 : Test started.
23:09:56 : 00/01  100.0%
23:10:37 : 00/02  100.0%
23:11:09 : 01/03  87.5%
23:11:21 : 02/04  68.8%
23:11:29 : 03/05  50.0%
23:11:45 : 04/06  34.4%
23:12:23 : 05/07  22.7%
23:12:33 : 05/08  36.3%
23:12:40 : 05/09  50.0%
23:13:07 : 06/10  37.7%
23:13:40 : 07/11  27.4%
23:14:08 : 08/12  19.4%
23:14:12 : Test finished.

----------
Total: 8/12 (19.4%)


Sounds good as the LAME 3.90.3 encode.
"I never thought I'd see this much candy in one mission!"

 
SimplePortal 1.0.0 RC1 © 2008-2021