Hi,

Questions:
1. Is it save to use the -V switches, with no corresponding preset (especially V 1 / V3)?
2. When is it more usefull, to use abr rather than -V n (e.g. --preset 120 better than -V 5)?
3. Are the displayed target average bitrates ok (+/- 10 kbps)?

Thanks

Edit: changed with Gabriel's suggestions. Added:-Y, -b, --lowpass, --resample
Edit: modified -b to fit with 3.96

Nice layout.


I'd say:
V2 - 190
V3 - 175
V4 - 165
V5 - 130

Is --r3mix still remapped to -V3 --vbr-new in 3.96 final?

Yes

I just want to pick up the first question of Vietwoojagig:

Is it save to use the -V switches, with no corresponding preset (especially V 1 / V3)?

The word you're looking for is safe, not save!

safe=not dangerous
save=keep or store

Don't flame me - I'm used to people with English as a second language at work getting irritated if I don't help them. Accepting that, with any language other than English I'm beyond help myself, so any help I give certainly isn't meant as criticism!

Sorry I can't answer the actual question! I guess that V1 can't be worse than V2, and V3 can't be worse than V4, so they're worth a try.

Cheers,
David.

Oh dear, shame on me. Of course I know the difference between save and safe. But sometimes I don't see the mistakes, even if you would nail them on my head.

Vietwoojagig,

don't worry. It's just a sign of your lingual mind starting to think phonetically. It's just transitioning from one set of phonemic transcription to another.

I make mistakes like that all the time. I wrote much better English when I knew and used it much less.

These days I'm bi-lingually handicapped. Sometimes I can't express myself properly in my native tongue (or my second native tongue), but English works out ok. Then again, often I think in Finnish and my English comes out really broken.

So, you're actually just advancing

And compared to the average American... don't even get me started

As for your questions, I don't think there is enough data to conclude 1) yet (at least I haven't found the answer). Well, I'm using -V switches on 3.96, based on some preliminary findings, but no conclusive data is available yet (?).

As for 2) and 3) I'm also interested. Perhaps the developers are the best ones to answer this, if they have time/interest.

cheers,
halcyon

PS It's an odd thing with English speaking natives: they complain when you try to speak their language and they complain when you don't. Nothing's ever good enough. And my God, what happens if you err on the side of correcting their English? A war breaks out, that's what happens ;-D

Well, considering the outdated nature of --r3mix and --preset medium being more recent and based off the proven --aps, I wouldn't be too surprised if -V4 (--vbr-any) could actually be better than -V3 --vbr-new, despite the lower bitrate. The question would be whether -V3 (--vbr-old) has more to do with --r3mix or with --aps -Y?

Only the name "--r3mix" is mapped to"-V3" and not "-V3" with the outdated code. So I'm quite sure that that V0 > V1 > V2 > V3 > V4 > ... > V9.

As I understand it, the names "r3mix", "preset medium", "preset standard", "preset extreme" are only mapped to V-switches for compatibilty reasons.

In future, I will only use the V-switches, knowing, that V2 is the setting for transparency.

From what I had understood, the -Vx now equal the --preset they are mapped with. So -V2 is the same as --preset standard and, under the same logic, -V3 --vbr-new would be --r3mix. Now you are telling me then that --r3mix is no longer --r3mix, but -V3 --vbr-new (whatever that is). Is this right? Has --r3mix been reduced to merely a symbolic link pointing to something entirely different?

Yes

I say by jove. Steady on old boy, what. Correcting the Queens - now that would be damn foolish move old bean. Spiffing what and pass the Pimms and Gordons.

PS. Yes I am English and even worse, a Cockney

Talk Cockney here or here!

Hmm... Interesting.
Gabriel, could you please confirm this? Or Vietwoojagig, could you show me where you know this from? Thank you.

Yes

Wow, that is really interesting. So the --r3mix tunings are really gone now (finally!) and all that remains is the name. This is truely significant!

I'd be interested in the answer to this question, too. No one has an idea?

-V2 (lame 3.95.1) gives me 141kbps on film track (pianist, russian translation). Double speed on 2.4 ghz
i'll do -V4 (lame 3.97alpha) this week-end on english track (in hope that picture quality will be better; using xvid).

Does LAME 3.96, -V2/--preset standard still use -b 128?

Yes. -b 96 did not work on some samples as desired.

Regarding "target bitrates" related to -V x, I'd like to point to this:
http://lame.sourceforge.net/lame_ui_example.html

In this sample UI, you can visually see the usual bitrate range of the different -V x levels

I have learned till now the difference between 'save' and 'safe' , but i still can't see here any clear answer on the question above, sorry.

So if I use "Lame.exe -V1 input.wav output.mp3"...
- ...will that produce smaller preset extreme mp3 files by using the 'extreme' maskings/shapings/modells but only a bit more aggressive?
- ...or will that produce bigger preset standard mp3 files by using the maskings/shapings/modells of preset standard but only a bit more gentle?
- ...or will '-V1' and '-V3' work beyond every well-tested preset?

Thanks for my enlightenment!

-V1 is between -V0 and -V2.
I do not know what to say more: it is not -V0 neither -V2, just -V1

...so -V1 includes the alt-preset code level tunings, but with a bitrate wetween standard and extreme.

I've removed the --r3mix line, since it doesn't mean --r3mix anyway.

No, there were no r3mix tunings. It was just a shortcut for a long string of command-line options that was advocated by some until it was beaten by (alt) preset standard.

sorry to reply to such old message

Thanks, that's just all i wanted to know!

I thought Gabriel was trying to move away from Dibrom's tuning (or "hack") by tuning in non-hack ways. In other words, hasn't he been removing more and more of Dibrom's tuning and trying to make it sound good by working from other angles? I don't understand source codes, so I can't be sure, but could someone (or Gabriel himself) comment on this? Sorry if I'm totally mistaken

No. My goal was to change the way those "tunings" are enabled. I wanted to be able to set each of those by separate parameters instead of a few "multi-purpose" parameters.

Would it be possible to add the -q switch to the tables? I believe --preset standard is -q 2 whereas --preset medium is -q 3.

Also, I am assuming there is no harm in say, --preset medium -q 0, since it theoretically may increase quality, even though it is much slower?

No

Rather than ask again - I just checked myself. --preset medium is qual=3, --preset standard is qual=3, and --preset extreme is qual=3 also according to --verbose command line when using 3.96.1. Interesting that even --preset insane is qual=3. It appears that -q 0/1/2 isn't used in any of the presets, unless I have something in EAC set wrong and it is adding other command line switches I don't know about.

What's with the q-value obsession? I'm not picking on you detokaal, I've been seeing it creeping up in other discussions. AFAIK, the -V settings were tuned using a given q-value and changing this isn't a reliable way to adjust quality. Rather just change the -V value. It seems the -V values have been distributed to cover the full bitrate spectrum, so it seems that anyone could find the ideal quality/space tradeoff for their particular ears and application.

Though I guess if you really want to be experimental, go for it. Might as well start tweaking the code as well. Perhaps Gabriel can either validate or invalidate this point.

Yes, changing -V is (generally) a better idea than changing -q.

Btw, I think that using the DEFAULT setting in current versions is fine. That mean that for vbr, you only adjust -V according to your needs and do not change anything else.

After reading through this thread, I'm still unclear (forgive me if I just don't GET IT). Is 3.96.1 -v5 a superior alternative to 3.90.3 --alt preset 128?

By superior, I mean superior sound quality for the targeted file size. Also, will the extra lowpass command used in Roberto's listening test improve -V5?

Since 3.96.1 -v5 was selected over 3.90.3 --alt preset 128 for Roberto's multi-format listening test, I ASSUME it is superior but I hate to assume anything in this forum.

Yes.

From the very start of this thread I'm still interested...

This is only a small part of the answer.

I just noticed that http://lame.sourceforge.net/USAGE now discourages the use of "-V 5" in favor of ABR 128. Where is Roberto's listening comparison test? I missed it and I'd like to read it. Thanks!

http://www.hydrogenaudio.org/forums/index....showtopic=21904

It is not "now". It's just that it hasn't been updated for a while.
note: I'd welcome an updated version

(From the best settings for 128kbps VBR thread.)



:edit: link

Hi.

I just registered myself for the first time ever to a forum of anykind

But this just seemed like a place of people who know the stuff they talk about,
and that's always good. Also this place has had by far the best attitude towards
"the new guy" making his first post, so bear with me, since this truly is my first post

So, here's my thing, I'm in a situation now to encode all of my audio cds
to mp3 format.

By judging form the discussions here, i really should use eac or plextools
for the ripping and lame (3.90.3 or 3.96.1 if i'm correct?) for the encoding.

Even more reading of this forum has led me to the conclusion to use vbr ape
or as you would put it in the format of -V0 while using lame 3.96.1.

So the thing I would like to know is about the q value,
is there any point in adjusting it upwards trough the commandline myself?

I mean seriously, I thought about this long time, that do I even dare to ask such a question,
because I respect what Gabriel has stated about the defaults,
but I just want to know that what are the potential effects of changing
the qval to 2 or even 1 or 0. What does it really change?
Does it affect the quality of the audio in any way?
I even read about it potentially lessening the outcoming quality of the finished "product",
is that still true?

well nothing more at this point, hope this didn't strike you guys as a totally stupid question.

thanks.

A bug regarding -q0 and -q1 has been recently identified and corrected in 3.97 alphas.
3.94-3.96.1 were affected by this problem. Regarding 3.90.3, I do not remember as it is too old.

Garbiel would this fix be backported to 3.96.x?

Probably not as we are not working on 3.96 anymore.

To get the average bit rate closer to 128 kbps, would you recommend"

V5

or--

with the switches used in the listening test--

V5 --athaa-sensitiivity 1 ??

What exactly does that --athaa-sensitiivity 1 mean, and what difference does it make over just V5?

use LAME 3.96.1 and -V5 --athaa-sensitivity 1. It results better qualits than -V5 (applies to -V5 only!) I've encoded 5 files or so and in average they were 128kbps. However, it is possible that some files have 111kbps and others 134kbps...Even if you get a bit over 128kbps in average, let's say 132kbps...who cares...it's not that much of a difference (in my example 0,5 kilobyte per second = 30 kilobyte per minute)...not a big deal, but the quality is the best mp3 has to offer in the 128kbps range...

my hearing is either getting really bad (i'm 26 years old!) or lame 3.96.1 is amazing because i cant abx -v6 --vbr-new

So I guess that means lame 3.96.1 @ -v6 --vbr-new will be my new portable choice.

Actually, the --athaa-sensitivity 1 switch was also recommended for -V4 on some thread I can't seem to find right now. I use it for my -V4 encodes and I haven't noticed any problems at all. (Actually, I can't ABX the difference, but if the golden ears say it's better, I trust them )

That is a valid question, one I bothered people with myself. -q 0 increases the quality of the psychoacoustic algorithm, it does not touch the bitrate at all.
In simple terms Lame will spend much more time examining each sample when encoding.

Lame 3.90.3 at --alt-preset standard is the recommended mp3 settings. Lame 3.97 final might replace 3.90.3 as the best encoder.


I use lame 3.96.1 -V 5 --athaa-sensitivity 1 -q 0, which might not improve the quality much over not using -q, but encodes fast enough for me and sounds good on my apex dvd player and my pc/stereo setup.

For cd's that I have to archive and won't have access to in the future, I use Musepack/MPC at --standard. MPC has fewer problem samples at higher bitrates and is tuned only for high transparent bitrates, but is not supported by any portables . It is very good for computer use.

preset standard is not the "absolute best" mp3. It's a compromize between transparency (I'd rather say 'robustness') and bitrate. For many people, --preset standard is not a good compromize: they could obtain the same level of transparency at lower bitrate with lower presets. For other people, --preset extreme improves the quality, and not only on killer samples.
For absolute best mp3, use --preset insane (or even freeformat at 640 kbps).