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: lame3995o (Read 40359 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

lame3995o

lame3995o is based on the fact that the lame3995n -Q1.7 quality is excellent to my ears and for all the music I tried except for the harp40_1 sample.
So it was a promising strategy for me to use the -Q1.7 quality parameters as a basis and optimize them for harp40_1 at -Q1.
I did that, created a new version lame3995o, and arrived at the following ABC/HR results:
  • lame3995n -Q1: 4.3
  • lame3995o -Q1: 4.5
  • lame3995o -Q0.5: 4.7

Compared to lame3995n lame3995o's different bahavior concerns the -Q values above 1.7. The main effect is between -Q1 and -Q0.5, The differences stop at -Q0.

You can download lame3995o from here.
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #1
Hi halb27.  I've been using -Q1 on my standard playlist for the past month, and have not noticed any quality issues under normal listening conditions.  As I get more time later in the summer, I will try a few ABC/HR tests on a few select tracks from my collection.  As a bonus, -Q1 uses less bitrate than standard LAME -V 1 does for my playlist.  Thanks again for your continuing attention and refinement to your fork of LAME.
                                                                 Best Regards,
                                                                                       LedHed8

Re: lame3995o

Reply #2
Thank you for testing lame3995o. I'm looking forward to your ABC/HR results.
Horst
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #3
I've noticed an audio distortion with lame3995o.
I ripped Billy Joel's song "Piano Man" and in the line "Sing us a song you're the piano man" the 'S' in sing distorts.

I'm using -Q0 setting in foobar. The CD is Billy Joel - The Essential Billy Joel 3.0

Re: lame3995o

Reply #4
[...] in the line "Sing us a song you're the piano man" the 'S' in sing distorts.
A little reminder:
Publishing your ABX tests's results (and if anyone would be interested, uploading a small sample of that passage) would be, as you'll probably recall, more in agreement with some of this community's TOS's - more specifically, TOS 8.
Listen to the music, not the media it's on.
União e reconstrução

Re: lame3995o

Reply #5
@soundping:
I also welcome if you could send a sound snippet of the critical spot so that I can have a look what's going wrong.
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #6
How to use lame3995o in foobar2000 ?

Re: lame3995o

Reply #7
You create a commandline encoder setting like this:

lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #8
thanks :)

Re: lame3995o

Reply #9
i will convert my files today with Q0.
With this bitrate (317-318 kps), it's better than cbr320 ?

Re: lame3995o

Reply #10
I guess for 99.99 percent of the music everything will be perfect no matter your choice.
I haven't tried recently but I guess the critical spot at sec.  ~3 of problem sample eig is in favor of lame3995o - Qx (it doesn't take - Q0).
Maybe there's a difference also for problem samples lead_voice or harp40_1. You can find out yourself. After all it's your ears that count.
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #11
I've noticed an audio distortion with lame3995o.
I ripped Billy Joel's song "Piano Man" and in the line "Sing us a song you're the piano man" the 'S' in sing distorts.

I'm using -Q0 setting in foobar. The CD is Billy Joel - The Essential Billy Joel 3.0

@soundping. Is the distortion you are hearing on the 1st chorus at about the 1:26.7 mark?  If so, I'm not able to hear a difference.  I have the same version from the 3.0 edition.  I've tried to abx on lower quality settings of 3995o, 3.99.5, and 3.98.4 versions down to V5 with no success.  In the past, I've been pretty sensitive to sibilant S distortions, so perhaps my hearing has worsened.  Then again, I hear an identical slight distortion in the 'S' in the FLAC and MP3 encodes.

Best regards, LedHed8

Re: lame3995o

Reply #12
I've noticed an audio distortion with lame3995o.
I ripped Billy Joel's song "Piano Man" and in the line "Sing us a song you're the piano man" the 'S' in sing distorts.

I'm using -Q0 setting in foobar. The CD is Billy Joel - The Essential Billy Joel 3.0

@soundping:
I also welcome if you could send a sound snippet of the critical spot so that I can have a look what's going wrong.

After all these days, will you kindly finally pass the baton of burden of proof, @soundping  , please?

EDIT:
To the admins/mods: I've just realized that neither halb27 nor me couldn't actually refer to soundping (member=96900) in our posts, since whenever we both typed "@soundping" in our replies, the editor has interpreted it as 'sound' (member=57038) - who is a totally different member - even when I at least, confirmed it by clicking on the right name in the drop-down list that pops out.

PS: Though it may be us doing something wrong, as @LedHed8  managed to do it properly. :-\


Listen to the music, not the media it's on.
União e reconstrução

Re: lame3995o

Reply #13
EDIT:
To the admins/mods: I've just realized that neither halb27 nor me couldn't actually refer to soundping (member=96900) in our posts, since whenever we both typed "@soundping" in our replies, the editor has interpreted it as 'sound' (member=57038) - who is a totally different member - even when I at least, confirmed it by clicking on the right name in the drop-down list that pops out.

PS: Though it may be us doing something wrong, as @LedHed8  managed to do it properly. :-\

You may mention members manually using the member=id# BBCode tag. For an example, quote the posts. It's a pain, but it works. I'll report this to ElkArte forum as a bug, or misfeature.

Re: lame3995o

Reply #14
I'm back online.

Tonight I'll make samples and post them.

Re: lame3995o

Reply #15
The original file is gone now. I re-ripped the track and the "S" distortion sound isn't there now.


Re: lame3995o

Reply #16
The original file is gone now. I re-ripped the track and the "S" distortion sound isn't there now.
Well, I'll be damned...
Listen to the music, not the media it's on.
União e reconstrução

Re: lame3995o

Reply #17
I made some test audio samples from 3995o.
The 30s source file is from a burn-in track sample.

Code: [Select]
https://mega.nz/#F!aEE3GQaa!kdDsaq5Pn4ycccPsZLBUwA
The source file is there and copies from settings -Q0, -Q0.5 and -Q1.

Re: lame3995o

Reply #18
I compared the flac with the -Q1 file and didn't notice a difference.
Is there an issue with one of the encoded files?
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #19
I didn't notice any problems. I was hoping someone else would test them out.

Re: lame3995o

Reply #20
I didn't notice any problems. I was hoping someone else would test them out.
Why bother if, as you were plain honest to admit, you couldn't detect anything wrong with said sample the second time around?

In my view, that was a simple case of claiming you'd found a problematic sample and then disproving it with (I hope) blind testing, afterwards.

Anyway, thanks for clarifying that.
Listen to the music, not the media it's on.
União e reconstrução

Re: lame3995o

Reply #21
The cutoff frequencies are not the same with Vx and Qx ?

Re: lame3995o

Reply #22
No they aren't. cutoff of -Qx is somewhat lower than that of -Vx.

The background for this is as follows:

-Vx internally uses a number of parameters controlling the accuracy of lossy encoding.
Lame has no real idea what a setting of these parameters has to be chosen in a specific situation for the music to be transparent.  That's why these parameters are static.
Sure these static parameters are carefully chosen so that usually music is fine even with -V5.
A higher quality -Vx setting means these parameters are chosen more restrictive so hat the deviation from the original is lower.  This way the probability of getting a transparent result increases. This is kind of a brute force quality management.

My -Qx quality setting works a bit different.
-Q1 for instance internally uses the accuracy parameters of -V3 as a quality basis. This gives the headroom for a dynamic quality management. For the known issues of very tonal problem samples as well as harpsichord music I derived criteria where audio quality is increased in a musical situation when lame3995o thinks this is necessary. This is a more selective approach than that of -Vx. (But don't think it is a very intelligent mechanism.  In order to have a strong effect on various problem samples I still had to do things in a pretty much brute force way).

I don't touch the lowpass anymore which is a function of -Vx, in case of -Q1 that of the underlying basic -V3.
If you prefer another lowpass, use the --lowpass setting.
lame3995o -Q1.7 --lowpass 17

Re: lame3995o

Reply #23
average kps of Q0 is 317 ...
So, better is -Q0 or CBR 320 ?

Re: lame3995o

Reply #24
Probability is close to zero that you would notice any difference.
Problem sample lead-voice is a promising candidate for finding a difference, maybe also harp40_1.
lame3995o -Q1.7 --lowpass 17