Skip to main content
Topic: Lame 3.98 beta 6 (Read 140402 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lame 3.98 beta 6

Reply #175
Forgive me for being a noob, but has anyone brought Martel's problem sample to the attention of Robert or another LAME developer?  I can also confirm the 'swoosh' sound in 3.98b6, and its absence in 3.90.1.

Lame 3.98 beta 6

Reply #176
What is all this noise ? Its low volume ringing. 3.97 is the same. I abx easily V5, V4 .. @ V3 I got a bit confused, but after a while I could do it. By V2 its really really close, but I managed 13/16.

Tracks with solo vocal, guitar are 'difficult'. Dibrom said this in old posts.
wavpack 4.8 -b3hx4cl

Lame 3.98 beta 6

Reply #177
Quote
Tracks with solo vocal, guitar are 'difficult'. Dibrom said this in old posts.

Yeah, but the LAME version from "Dibrom's time" doesn't appear to have this particular problem when 128 or 160 kbps CBR is used. (I didn't try VBR because Martel's report was about CBR.)

I hope the regression is caused by some kind of mishap and can be easily fixed without breaking something else.

Lame 3.98 beta 6

Reply #178
@halb27

does the problem you spotted in LAME persists in 320kbps CBR?

I've tried to build an uptodate system based on my old one, but didn't succeed. Looks like the mainboard or memory (or their cooperation) isn't alright. So I reinstalled my old components.

I just relistened to the problem again and - shame on me - can't abx it at all using -V0. Maybe I'm still too upset with my system trouble. I also tried -b320 and got at 5/6 which in the end turned out to be 6/10 - the very kind of result I formerly often got with french woman singers: too good to be ignored and too bad to take it as a proper abx result. Trying -V0 again for a final check I had no chance to abx it.

Despite my poor results I think it's enough to conclude that -b320 doesn't solve the problem.
lame3995o -Q1.7
opus --bitrate 140

Lame 3.98 beta 6

Reply #179
Quote
@halb27

does the problem you spotted in LAME persists in 320kbps CBR?


I have the same experience as halb27. fail at 320kbps.

But even at q-0 the problem is very slight and not annoying. (for me at least) It may actually transparent to your ear.

Lame 3.98 beta 6

Reply #180
Why there is no sources for 3.98 betas on SourceForget.net: LAME Dowload page?

There was betas before 3.97 version. Or its not good enought for it?


Lame 3.98 beta 6

Reply #182
It has been some time since I followed LAME development and reading this thread shows that there are regressions and improvements provided by this new beta version, however I am not into the recent development/testing process.

I have to do some encodes at the V5 range and I was wondering if 3.98b6 is safe to use or if I should use the stable 3.97 version instead. I intend to use --vbr-new as well to speed up things. Do you see any problems?

thanks

Lame 3.98 beta 6

Reply #183
... I have to do some encodes at the V5 range and I was wondering if 3.98b6 is safe to use ....

I'm very happy with 3.98b6 but my experience is with -V3+, and with this I prefer 3.98 over 3.97. I guess the advantageous behavior is valid for -V5 too, but can't really tell about it.
lame3995o -Q1.7
opus --bitrate 140

 

Lame 3.98 beta 6

Reply #184
I tested a 30 second sample using 3.98b6 (Mar 12, 2008) vs 3.97 both with -V 6 --vbr-new, and on this sample, I noticed regression from 3.97. I'm sorry that I haven't done any listening test on Lame lately, and this finding may be too late to be any help for the final 3.98 release, but it might amount to something for your future development. I will try to find more regression later if I can.

Considering that very few people (or none that I remember) have reported regression issues with 3.98b6 at low bitrate, this may well be an exception, but I don't think this sample is a killer sample or anything. I will upload it to the Upload section in a few minutes (first 30 seconds of Giant Steps by Coltrane).



ABC/HR Version 0.9b, 30 August 2002
Testname: 3.97 -V 6 --vbr-new vs 3.98b6 -V 6

1R = C:\My Test Samples\gsteps\398.mp3.wav
2L = C:\My Test Samples\gsteps\397.mp3.wav

---------------------------------------
General Comments:
sample #2 was much harder to ABX.
---------------------------------------
1R File: C:\My Test Samples\gsteps\398.mp3.wav
1R Rating: 1.0
1R Comment: Distortion is severe throughout the sample.
---------------------------------------
2L File: C:\My Test Samples\gsteps\397.mp3.wav
2L Rating: 2.0
2L Comment: Distortion is not nearly as severe as #1.
---------------------------------------
ABX Results:
Original vs C:\My Test Samples\gsteps\398.mp3.wav
    8 out of 8, pval = 0.004
Original vs C:\My Test Samples\gsteps\397.mp3.wav
    7 out of 8, pval = 0.035
C:\My Test Samples\gsteps\398.mp3.wav vs C:\My Test Samples\gsteps\397.mp3.wav
    8 out of 8, pval = 0.004

Lame 3.98 beta 6

Reply #185
Just an observation from your results:

sample #2 was much harder to ABX.

1R Rating: 1.0

2L Rating: 2.0
2L Comment: Distortion is not nearly as severe as #1.

ABX Results:

Original vs C:\My Test Samples\gsteps\397.mp3.wav
    7 out of 8, pval = 0.035
C:\My Test Samples\gsteps\398.mp3.wav vs C:\My Test Samples\gsteps\397.mp3.wav
    8 out of 8, pval = 0.004



I would have no problems ABX ing something that I rate as "Annoying".

Lame 3.98 beta 6

Reply #186
Quote
' date='Mar 24 2008, 05:32' post='554570']
Just an observation from your results:



sample #2 was much harder to ABX.

1R Rating: 1.0

2L Rating: 2.0
2L Comment: Distortion is not nearly as severe as #1.

ABX Results:

Original vs C:\My Test Samples\gsteps\397.mp3.wav
    7 out of 8, pval = 0.035
C:\My Test Samples\gsteps\398.mp3.wav vs C:\My Test Samples\gsteps\397.mp3.wav
    8 out of 8, pval = 0.004



I would have no problems ABX ing something that I rate as "Annoying".


That means that my ears aren't that good.

Lame 3.98 beta 6

Reply #187
If people are complaining about 3.98b6, why aren't people using the previous beta that apparently didn't have those issues?

Is it because Rarewares doesn't have it anymore?

- Spike

Lame 3.98 beta 6

Reply #188
That means that my ears aren't that good.


It is not about ears. It is about terminology. If it annoys you, you're disturbed while listening to it. i.e. you wouldn't listen to it for a prolongued period of time.

I am sure you meant to rate it between 4 and 5, which stands for "Perceptible, but not annoying". I.e. You could happen to hear it, (in your case you had to put efforts on it), so in a casual listening you wouldn't mind/notice it, and as such, not annoy you.

I stress this point, because else, the listening tests that we run sometimes here, wouldn't have much sense. If a thing is either 5 (Not perceptible) or 2 (Annyoing), The scale is converted to something like is either white or black.

Don't take this as personal, but rather as a thing to think about when doing abc-hr tests, alright?

Lame 3.98 beta 6

Reply #189
The file I promised has been uploaded here.

Lame 3.98 beta 6

Reply #190
I can confirm the problem. 3.97 is better here with -V6 for me too .
lame3995o -Q1.7
opus --bitrate 140

Lame 3.98 beta 6

Reply #191
is there a release schedule for the final version of 3.98?

Lame 3.98 beta 6

Reply #192
Just noticed Updated Beta 6 :

"LAME 32bits version 3.98 (beta 6, Mar 12 2008) (http://www.mp3dev.org/)"

Sources - http://sourceforge.net/project/showfiles.p...;package_id=309

Binaries - http://rarewares.org/mp3-lame-bundle.php

"lame 3.98 beta 6 (small fix) Notes(2008-03-12 13:47)"

CVS - http://lame.cvs.sourceforge.net/lame/lame/...me/?sortby=date

// @ devs : check my ticket please

Lame 3.98 beta 6

Reply #193
Just noticed Updated Beta 6 :

"LAME 32bits version 3.98 (beta 6, Mar 12 2008) ...
Wouldn't it be better to have a new beta version instead of multiple versions of the same beta?  This causes confusion when talking about found bugs.

Lame 3.98 beta 6

Reply #194
Anyone know what's actually changed in the new beta 6?

- Spike



Lame 3.98 beta 6

Reply #197
The changelog on the url mentioned above still says, "LAME 3.98 beta 6  December 16 2007". Has it been updated to incorporate the changes that went into the subsequent re-released beta6?

--Vinu.


You can read http://lame.cvs.sourceforge.net/lame/lame/...me/?sortby=date and other pages.

Code: [Select]
psymodel.c              1.178   9 hours     robert     some simpler spreading function for VBR NEW
set_get.c               1.87    9 hours     robert     some simpler spreading function for VBR NEW
version.h               1.114   9 hours     robert     some simpler spreading function for VBR NEW
presets.c               1.65    9 hours     robert     some simpler spreading function for VBR NEW
lame_global_flags.h     1.44    9 hours     robert     some simpler spreading function for VBR NEW
id3tag.c                1.52    19 hours    robert     changing some links from "www.mp3dev.org/mp3" to "www.mp3dev.org" and from "www....


Interesting: current CVS compile says it's beta 7
Code: [Select]
LAME 32bits version 3.98 (beta 7, Apr  6 2008) (http://www.mp3dev.org/)


Edit: formatting

Lame 3.98 beta 6

Reply #198
Time for a new thread (click) then 
Let's not get into the situation we had with b6 that the release number had changed but the beta wasn't ready for release yet. It's official now
Anyone know what's actually changed in the new beta 6?
could it be one of fixes in the remarks on rarewares?
Quote
(Includes the decoding fix for mp3 files and the fix for decoding through stdout. Both now included in CVS. Also now includes Nyaochi's Album Art Id3 Patch, also in the CVS.)
In theory, there is no difference between theory and practice. In practice there is.

Lame 3.98 beta 6

Reply #199
I'll get some beta 7 compiles up in a couple of hours, or so. (I was away over night and only just returned home.  )

 
SimplePortal 1.0.0 RC1 © 2008-2020