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: Lame versions and alphas - READ THIS! (Read 53392 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lame versions and alphas - READ THIS!

Reply #25 –
OK, i may be talking bollox here, but i seem to recall a version of lame 3.9.2 with "nonlinear quant scale"... i may be confused with TMPGenc, because i was linked to both through doom9's download page

i know that my lame 3.9.2 is giving me much better very-low bitrate encodes than my other installed lame compiles.... does anyone know about this?  (i've searched, but not as hard as maybe i could)

Lame versions and alphas - READ THIS!

Reply #26 –
Quote
OK, i may be talking bollox here, but i seem to recall a version of lame 3.9.2 with "nonlinear quant scale"... i may be confused with TMPGenc, because i was linked to both through doom9's download page

i know that my lame 3.9.2 is giving me much better very-low bitrate encodes than my other installed lame compiles.... does anyone know about this?  (i've searched, but not as hard as maybe i could)

I think you're thinking of DSPguru's custom lame_enc.dll with a non-linear psychoachoustic model, from here: http://dspguru.notrace.dk/lame_encNL.zip

Lame versions and alphas - READ THIS!

Reply #27 –
WORD!  thanks.

hehe.. i was confused with an update of TMPGenc where the quantscale was changed (in a similar manner it would appear).

hmm.. if that's a DLL though, that means there must be some significant changes in lame 3.95.1 in the low bitrate areas (96 CBR).  interesting.

this is sort of just curiosity for me - seeing how low i can go before LAME breaks.

[edit]

any idea what version of lame this dll is?  doesn't seem to say on dspguru's site

Lame versions and alphas - READ THIS!

Reply #28 –
Quote
WORD!  thanks.

hehe.. i was confused with an update of TMPGenc where the quantscale was changed (in a similar manner it would appear).

hmm.. if that's a DLL though, that means there must be some significant changes in lame 3.95.1 in the low bitrate areas (96 CBR).  interesting.

this is sort of just curiosity for me - seeing how low i can go before LAME breaks.

[edit]

any idea what version of lame this dll is?  doesn't seem to say on dspguru's site

3.92, so it's one of the decent ones!  Although with the amended psychoachoustic model that comment may be entirely inappropriate!!

Lame versions and alphas - READ THIS!

Reply #29 –
Quote
Although with the amended psychoachoustic model that comment may be entirely inappropriate!!


i hear you...  but i always loved to tinker with things, so for me the more versions to play with the better.

that said, i do all my archival encodings in --aps... with the occasional -Y as i can't ABX it on actual music, rather than problem samples

you guys would hate to see the command lines i play with just for fun.  more to learn just what they do rather than try to improve a preset of course.  though i've been surprised that commandline tinkering can produce better results for VERY low bitrates.  i'll post a sample if anybody really cares...

Lame versions and alphas - READ THIS!

Reply #30 –
I care.

I'd like to see something that could beat "--alt-preset 96" at it's own game. For me artifacts are very easy to hear at this level.

Lame versions and alphas - READ THIS!

Reply #31 –
Quote
I care.

I'd like to see something that could beat "--alt-preset 96" at it's own game. For me artifacts are very easy to hear at this level.

Somehow I don't think 96kbit/s is what MugFunky meant when he said " . . . VERY low bitrates."

edit:  To me 96kbit/s is a mid-range bitrate.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame versions and alphas - READ THIS!

Reply #32 –
Deeswift, I know you last posted to this thread in August, and the last time anyone posted to it was in February, but in case you're still interested...

You asked if there was a LAME setting in between --alt-preset standard and --alt-preset extreme - one that creates files with better quality than the former, and with smaller file sizes than the latter. Well, I'm using LAME 3.95.1, and after a bit of experimentation, I've found that apparently --alt-preset standard is equivalent to -m j -V 2, and --alt-preset extreme is equivalent to -m j -V 0.

Therefore, there is a setting that meets your needs: -m j -V 1. I've tried it, and the q level, lowpass filter settings, and minimum bitrate are indeed right between those of --alt-preset standard and --alt-preset extreme. The resulting file size is right in the middle as well.

Personally, I'm one of those slightly compulsive types who can't be satisfied even with --alt-preset extreme, so when I'm encoding wave files to MP3, I use --alt-preset extreme -qval=0, in spite of the fact that there have been multiple posts here recommending that users not add command line switches to the LAME presets - and also in spite of the fact that I've searched high and low and have found no explanation anywhere of what the heck "qval" is, or exactly what parameters it controls! Hey, if there's any chance that it makes MP3s sound better, then give me the best "qval" there is, baby!

Lame versions and alphas - READ THIS!

Reply #33 –
Quote
I use --alt-preset extreme -qval=0, in spite of the fact that there have been multiple posts here recommending that users not add command line switches to the LAME presets - and also in spite of the fact that I've searched high and low and have found no explanation anywhere of what the heck "qval" is, or exactly what parameters it controls! Hey, if there's any chance that it makes MP3s sound better, then give me the best "qval" there is, baby!

THere is also more than a passing chance that it makes things sound worse, too.

Lame versions and alphas - READ THIS!

Reply #34 –
Quote
THere is also more than a passing chance that it makes things sound worse, too.

Explain, please!

You might possibly be right, but as a test, I encoded the same file using --alt-preset extreme and --alt-preset extreme -qval=0. The one with the -qval=0 switch took more than four times as long to encode, and came out about two tenths of a percent smaller.

Granted, this doesn't prove anything, but it suggests that perhaps the preset with the added qval switch bears the same sort of relationship to the "plain" preset that the "plain" preset has to its "fast" equivalent - that is, it takes longer and gives a more "careful" result.

Lame versions and alphas - READ THIS!

Reply #35 –
Quote
Well, I'm using LAME 3.95.1, and after a bit of experimentation, I've found that apparently --alt-preset standard is equivalent to -m j -V 2, and --alt-preset extreme is equivalent to -m j -V 0.

IIRC -V 0 was remapped to --preset extreme and -V 2 was remapped to --preset standard, in version 3.95.1.  The -m j is redundant, they are entirely the same.  Whether -V 1 offers any quality improvment to --preset standard, or not, I don't know.  Perhaps a LAME dev could enlighten us.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame versions and alphas - READ THIS!

Reply #36 –
Quote
The -m j is redundant, they are entirely the same.

I don't use -m j with the --alt-presets (or "--presets"; yes, I've noticed through trial and  error that, as you point out, --preset extreme is the same as --alt-preset extreme). I was just trying to make the point that -m j -V 0 seems to produce the same results as --alt-preset extreme or --preset extreme.

In fact, I just did a test, encoding the same file using both -m j -V 0 and --alt-preset extreme. The resulting file sizes were exactly the same, which would strongly suggest that both files are exactly the same.

Quote
Whether -V 1 offers any quality improvment to --preset standard, or not, I don't know. Perhaps a LAME dev could enlighten us.

A LAME developer would, indeed, certainly be able to offer more authoritative insights on this matter than can I. However, if --alt-preset standard's -V 2 sounds better than --alt-preset medium's -V 4, and --alt-preset extreme's -V 0 sounds better than --alt-preset standard's -V 2, then doesn't it stand to reason that -V 1 would also sound better than -V 0, albeit perhaps to an insignificant degree?

Lame versions and alphas - READ THIS!

Reply #37 –
Quote
3.94 Alphas and ALL future alpha versions -> ALPHA versions are for testing only! These include new code and are not tested or tuned at all. Many things are usually seriously broken or changed! DO NOT USE for anything, except for testing!

Could someone tell me if there's anywhere where all versions of Lame are available for download?  This thread started to document 3.90.3 and post 3.90.3 versions a year ago.  But up front it doesn't describe versions 3.4, 3.5, 3.95.1 and 3.6 that I've read discussions on, or provide a link to where they can be downloaded.  They're only discussed to some degree in later posts to this thread.

The most discussed versions seem to be here: http://www.rarewares.org/mp3.html 

Is that an 'official' HA download site?  Is there somewhere else where ALL of the various Lame compiles are listed, described, and available for download?
Geopoliticus Child Watching the Birth of the New Man

Lame versions and alphas - READ THIS!

Reply #38 –
Is this still a useful thread, seeing as how the latest version is LAME 3.96.1?  I don't think I've seen too many alpha binaries distributed since, either.

Lame versions and alphas - READ THIS!

Reply #39 –
I updated the post to make it useful again.
Thanks for pointing me to this.
"To understand me, you'll have to swallow a world." Or maybe your words.

Lame versions and alphas - READ THIS!

Reply #40 –
Quote
Whether -V 1 offers any quality improvment to --preset standard, or not, I don't know. 
[{POST_SNAPBACK}][/a]


I have found that for me it does offer improvement over both 3.90.3 and 3.96 aps and ps, respectively (see [a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=20715]this thread[/url]).

In ABX testing in the above mentioned thread, I realized that aps and ps are not only not transparent on certain samples, but glaringly obvious in some. I've found that -V1 improved the artifacts enough to the point that they were slight and not nearly as annoying. I have decided it is the best compromise for me between quality and size, and I now currently use 3.96.1 -V 1, which replaced my previous 3.90.3 aps setting.

I think its great that we have this new option between ps and pe.