Skip to main content

Topic: lame3100k - bringing constraint VBR to Lame (Read 35992 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • [JAZ]
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #50
So, in this case, it looks like it's "simply" a lossy sample used in the making of the song  ( Yes , sometimes that is done ).
Especially in the 60second spectogram this is the most plausible reason (keeps the same shape while top frequency is higher on other sounds)

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #51
...This time did only 8 tests, but I ended up scoring the same results in both.
Hope it will help? ...

Thanks a lot, yes it does.
These results are in line with what you found when comparing Lame3.100a2 -Vn with lame3100k -Vn --cvbr auto in general.
So no special problem with -V1 --cvbr auto.
lame3995o -Q1

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #52
I'll see what I can do, but sorry at the moment I am busy with other things.
lame3995o -Q1

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #53
Finally I've managed to finish version 3.100l.
It has all the stuff mentioned in this thread.
In order to allow for final changes due to user experience I give it here before starting a new thread for it.

You can download Lame3100l from here.

What’s lame3100l?

It's an extension to Lame 3.100 alpha2 which offers constraint vbr mode.
-Vn defaults to this constraint vbr mode. Average bitrate increase for quality levels -V5 to -V0 is only 2 kbps (for a variety of pop music) compared to original Lame.
Audible advantage of this variant is rare. It can be experienced most easily with samples eig and lead-voice.

--cvbr x, 9 >= x >= 0 or x = -1
Constrained vbr mode is controlled by cvbr level.
Lowest cvbr level is 9.0, highest cvbr level is 0.0.
Default cvbr level for -V5 (and lower) is 9.0, -V4 defaults to cvbr level 8.0, ..., and -V0's default cvbr level is 4.0.
Using the --cvbr x option overrides the default cvbr level with cvbr level x.
Using a cvbr level better than 4.0 together with -V0 (or -V1) can improve those small issues which can still be there when using -V0. You can experience this for instance with problem sample harp40_1.
-V0 --cvbr 0 (yielding 316 kbps on average) is an extremely high quality alternative to cbr 320.
--cvbr -1 turns cvbr mode off and falls back to original Lame behavior.

--bCVBR x, 135 <= x  <= 316
x is the desired average bitrate for a variety of pop music.
Instead of using -Vn and trying different n's to arrive at the desired bitrate, this option can be used.

For more details see the provided documentation.
lame3995o -Q1

  • Kamedo2
  • [*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #54

Thank you for the update. Now it's much easier to use. I'm going to provide more detailed analysis in August. I'm too busy now.

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #55
I just changed the --bCVBR x behavior for 230 < x < 278 a bit. It has a certain bias now towards a higher cvbr level. This has a positive effect on the remaining tiny issues of herding_calls and harp40_1.

@goa pride: I tried to produce a 64 bit version. Would have to install the Windows SDK to do so. I tried several times but didn't succeed in installing it - I always get an error message which isn't informative to me. Sorry.
  • Last Edit: 24 July, 2013, 01:27:28 PM by halb27
lame3995o -Q1

  • halo001
  • [*]
lame3100k - bringing constraint VBR to Lame
Reply #56
Great! This version got both options for tuning and simplicity. Very nice.

  • Kamedo2
  • [*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #57
Average bitrates of many pops and jazz albums(44.1k stereo).

  • SubV
  • [*]
lame3100k - bringing constraint VBR to Lame
Reply #58
halb27
Have you tried to increase the lowpass value for high bitrates (i.e. -V0 --cvbr 0)? Since 18200 Hz seems a little bit low to me.

I wonder how increasing the lowpass to 19 kHz will affect the perceived quality?

  • db1989
  • [*][*][*][*][*]
  • Global Moderator
lame3100k - bringing constraint VBR to Lame
Reply #59
Without results showing that you can detect frequencies above 18 kHz in any context other than a sine wave playing at an extremely loud and musically irrelevant SPL, I doubt they have any impact on perceived quality whatsoever.

  • goa pride
  • [*][*]
lame3100k - bringing constraint VBR to Lame
Reply #60
i used -V0 --cvbr auto --lowpass 23 with lame3100k
i get error with this new lame3100l version!
Lame 3100a [32bit and 64bit]                           https://bit.ly/1VPTGbO
WinMP3Packer all-in-one [32bit and 64bit]     https://bit.ly/1PAtMS1
MP3Suite.bat customizable [64bit encoders]   https://bit.ly/1ZVOVxx

  • mezenga
  • [*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #61
The default lowpass values were already discussed at lame3100g´s thread [a href='index.php?act=findpost&pid=817295']here[/a] and [a href='index.php?act=findpost&pid=817351']here[/a].

Anyone with a different personal preference can use the lowpass switch to override the default values.

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #62
Have you tried to increase the lowpass value for high bitrates...?

No, and I won't, for the reasons given by db1989 and mezenga.
  • Last Edit: 25 July, 2013, 04:18:04 PM by halb27
lame3995o -Q1

  • BFG
  • [*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #63
If you guys are really that worried about the lowpass filter, you could always try --lowpass -1 -Y.  That's what I use and I've been very satisfied with it.

EDIT: Halb, I forgot to say thank you, once again, for version L!  I continue to be impressed with the results, and from a casual listening perspective, I'm convinced your versions of LAME represent the best MP3 encoders available.
  • Last Edit: 27 July, 2013, 01:54:20 AM by BFG

  • halo001
  • [*]
lame3100k - bringing constraint VBR to Lame
Reply #64
I agree with BFG. I've done test on my several FLAC music collections and I'm surprised I could not ABX the tracks using my settings -V0 --cvbr 3 --lowpass 19916. I'm very convinced that this version in this case eliminated most of the audible issues.
  • Last Edit: 27 July, 2013, 02:49:51 AM by halo001

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #65
Thanks for all the flowers.
lame3995o -Q1

  • goa pride
  • [*][*]
lame3100k - bringing constraint VBR to Lame
Reply #66
-V0 --cvbr auto --lowpass 23 do not work!, how come?
-V0 --cvbr 3.5 --lowpass 23 WORK!

is --cvbr auto removed?
  • Last Edit: 27 July, 2013, 06:39:05 AM by goa pride
Lame 3100a [32bit and 64bit]                           https://bit.ly/1VPTGbO
WinMP3Packer all-in-one [32bit and 64bit]     https://bit.ly/1PAtMS1
MP3Suite.bat customizable [64bit encoders]   https://bit.ly/1ZVOVxx

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #67
Yes, --cvbr auto is removed.
It's not necessary any more because -Vn defaults to --cvbr auto mode.
See the docs.
  • Last Edit: 27 July, 2013, 06:44:20 AM by halb27
lame3995o -Q1

  • [JAZ]
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #68
If you guys are really that worried about the lowpass filter, you could always try --lowpass -1 -Y.  That's what I use and I've been very satisfied with it.

EDIT: Halb, I forgot to say thank you, once again, for version L!  I continue to be impressed with the results, and from a casual listening perspective, I'm convinced your versions of LAME represent the best MP3 encoders available.



I agree with BFG. I've done test on my several FLAC music collections and I'm surprised I could not ABX the tracks using my settings -V0 --cvbr 3 --lowpass 19916. I'm very convinced that this version in this case eliminated most of the audible issues.



Do any of you both have evidence that using vs not using the lowpass setting can be passed with an ABX test? Also, BFG, with which settings do you use that? (I hope you are aware that it is only useful with V0..2, because the rest of VBR, CBR and ABR already have -Y enabled)
  • Last Edit: 27 July, 2013, 07:37:30 AM by [JAZ]

  • ash92
  • [*]
lame3100k - bringing constraint VBR to Lame
Reply #69
can you give setting Line options  for EAC V1.0 beta 3.    I want the best quality Auto .Sorry for the mistakes, I use Google Translate

  • halb27
  • [*][*][*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #70
For best quality use
--bCVBR 316 --noreplaygain --silent %source% %dest%
as the Lame commandline from within EAC.
I suggest however to try a somewhat lower --bCVBR value to see whether that works for you.
lame3995o -Q1

  • ash92
  • [*]
lame3100k - bringing constraint VBR to Lame
Reply #71
halb 27 - Thank You !

  • LedHed8
  • [*][*]
lame3100k - bringing constraint VBR to Lame
Reply #72
halb27, once again, thank you for all your hard work on your lame extension.  To my ears, it produces the finest and most pleasant sounding mp3's in existence.  I suspect that your V2 auto setting (cvbr6) is overkill for 99.9% of all music.  I hope to find some time in August to conduct a small abx test with a particular album that I have had successful abx results with in the past.  On this particular album, I've been able to abx certain Lossless tracks against iTunes/qaac 192 cvbr, Std lame V2, and Vorbis q6.  I'm rather hopeful that your lame extension will finally put "my problem" album to rest for me with an encoder setting around 200 kbps so my mind can peacefully allow me to stop using overkill settings. 

Best regards.

  • BFG
  • [*][*][*]
lame3100k - bringing constraint VBR to Lame
Reply #73
Do any of you both have evidence that using vs not using the lowpass setting can be passed with an ABX test? Also, BFG, with which settings do you use that? (I hope you are aware that it is only useful with V0..2, because the rest of VBR, CBR and ABR already have -Y enabled)

I do not...my hearing is not very good, and I can rarely ABX -V3 MP3s versus FLACs .  I also cannot perceive any sound over approximately 15,500Hz.  That said, I do try to encode MP3s at a higher quality than what I need, in case I have friends or family listening.  I was aware that -Y is used with -V3 and lower; I always use -V0 so have to specify it.

My theory (which I do not believe can be proven or disproven, as it is a qualitative statement and depends on many variables) is that an inaccurately modeled full range is superior to an accurately modeled limited range, if the difference is only in the very high frequencies (>16kHz).

Finally, I will say this much: I agree with either using -Y or a lowpass.  I really don't see any reason to accurately model the full 22,050Hz range; it's better to reserve those bits for the range that is perceivable.

  • db1989
  • [*][*][*][*][*]
  • Global Moderator
lame3100k - bringing constraint VBR to Lame
Reply #74
To my ears, it produces the finest and most pleasant sounding mp3's in existence.


Quote
I suspect that your V2 auto setting (cvbr6) is overkill for 99.9% of all music.  I hope to find some time in August to conduct a small abx test with a particular album that I have had successful abx results with in the past.
Until you have objective evidence that this mod performs better than the others at producing transparency in a double-blind test, we are not interested in unsubstantiated suspicions about how it sounds, which violate TOS #8.