Best version of LAME and best arguments...

Topic: Best version of LAME and best arguments... (Read 36113 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

Best version of LAME and best arguments...

2005-03-21 22:04:27

Is there a recognised best version of LAME out there that the audiophiles among you use to compress their tracks? I'm looking to compress some music using LAME, and I don't care how long it takes (so -q 0 is fine with me, for example). I'd like the resulting music to be the best MP3 quality possible, but with VBR.

Any ideas?

Best version of LAME and best arguments...

Reply #1 – 2005-03-21 22:14:02

Quote

Is there a recognised best version of LAME out there that the audiophiles among you use to compress their tracks? I'm looking to compress some music using LAME, and I don't care how long it takes (so -q 0 is fine with me, for example). I'd like the resulting music to be the best MP3 quality possible, but with VBR.

Any ideas?
[{POST_SNAPBACK}][/a]

There is, Its Lame 3.90.3
It has been tested allot and --alt-preset standard is a good setting.

Please read

[a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=28124]Lame settings[/url]

The recommended version and where to download it.

and last but not least

FAQ --- Lots of good info
[span style='font-size:8pt;line-height:100%'] edit: spelling [/span]

Best version of LAME and best arguments...

Reply #2 – 2005-03-21 22:16:28

yeah, it's under the pinned "recommended ..." threads in MP3 section.

so...
recommended version: 3.90.3
recommended parameter: --preset standard

later

edit: oh, and i type slow evidently

Best version of LAME and best arguments...

Reply #3 – 2005-03-21 23:51:19

Thanks for the info! I have already tried --alt-preset extreme, but I was wondering why it used -q 2 instead of -q 0. In my naivete, I assumed -q 0 was better - is this not always the case?

Best version of LAME and best arguments...

Reply #4 – 2005-03-22 00:05:29

No it isn't. The best setting is --preset-standard.

you shouldn't worry about other switches if you want the best quality possible...

Best version of LAME and best arguments...

Reply #5 – 2005-03-22 00:07:25

Quote

Thanks for the info! I have already tried --alt-preset extreme, but I was wondering why it used -q 2 instead of -q 0. In my naivete, I assumed -q 0 was better - is this not always the case?
[a href="index.php?act=findpost&pid=284306"][{POST_SNAPBACK}][/a]

well, it is true for CBR files.
for VBRs, -q0 results in smaller files..

someone correct me if I am wrong..

but me personally, I use LAME 3.96.1 -V2 -q0

btw, you may use --scale x.xx if you use replaygain.. it saves you some bits, you may rather use for music..

Best version of LAME and best arguments...

Reply #6 – 2005-03-22 03:13:27

^^

-q0 and q1 are buggy in 3.96.1

More info:

http://www.hydrogenaudio.org/forums/index....ic=18091&st=25#

http://www.hydrogenaudio.org/forums/index....showtopic=11925

Best version of LAME and best arguments...

Reply #7 – 2005-03-22 04:11:16

-q 0 and -q 1 were screwed up in 3.93.1 to 3.96.1, though in earlier versions it still sounded badder than -h.I would only use these switches as of lame 3.97 alpha 7.

How does -q 0 result in a smaller file I thought it was slower because it used a different algorithm which is supposed to achieve better quality at the selected bitrate.

Best version of LAME and best arguments...

Reply #8 – 2005-03-22 07:16:08

Quote

Is there a recognised best version of LAME out there that the audiophiles among you use to compress their tracks?
[a href="index.php?act=findpost&pid=284271"][{POST_SNAPBACK}][/a]

None. 3.90.3 is not recommanded because quality is superior (it's nevertheless superior to lame 3.89), but for another reason (it was tested... 40 months ago). I've recently tested lame 3.90.3 vs 3.97 alphas at high VBR settings, and lame 3.90.3 quality was inferior on average. Currently, nobody sent any proof or evidence that lame 3.90.3 --preset-standard or --preset extreme is superior to latest stable release (3.96.1) or alpha one (3.97 alpha). People claiming any superiority or saying that 3.90.3 is "better" without valid listening tests are infringing HA Term of Service (#8). 3.90.3 is only tested (which doesn't mean anything nowadays), not superior. That's all.

Best version of LAME and best arguments...

Reply #9 – 2005-03-22 07:30:02

Quote

No it isn't. The best setting is --preset-standard.

you shouldn't worry about other switches if you want the best quality possible...
[a href="index.php?act=findpost&pid=284308"][{POST_SNAPBACK}][/a]

Does this mean that --alt-preset extreme is worse quality than --alt-preset standard, even though the average bitrate is lower? That doesn't make sense to me, although I'll admit that not everything does.

Best version of LAME and best arguments...

Reply #10 – 2005-03-22 07:34:38

Quote

Does this mean that --alt-preset extreme is worse quality than --alt-preset standard, even though the average bitrate is lower? That doesn't make sense to me, although I'll admit that not everything does.

He was talking about -q2 and -q0

Best version of LAME and best arguments...

Reply #11 – 2005-03-22 07:35:39

The best setting is --preset insane. The ultra-best one is probably -b640 --freeformat.

--preset standard is considered as the most efficient (apparently transparent for most people). --preset extreme is better, at least on some samples, but higher bitrate too. For most people, switching from -V2 (standard) to -V0 (extreme) is not worth.

Best version of LAME and best arguments...

Reply #12 – 2005-03-22 08:33:27

@guruboolez
You're giving 3.90.3 the hard pressure *grin* I won't complain.
Still, even if you're factually right, i think it is the wrong time to do this. Save your energy for pressure on 3.90.3 until 3.97beta is out. Trying to push an alpha build may in the long run be more damaging than do good.

Just my humble opinion.
- Lyx

edit: this is an issue of habbit and psychology with terms(alpha, beta, stable), not a quality one. Still, this can be important as building trust on the term "ha.org recommended".

Best version of LAME and best arguments...

Reply #13 – 2005-03-22 08:40:16

Quote

@guruboolez
You're giving 3.90.3 the hard pressure *grin* I won't complain.
Still, even if you're factually right, i think it is the wrong time to do this. Save your energy for pressure on 3.90.3 until 3.97beta is out. Trying to push an alpha build may in the long run be more damaging than do good.

Just my humble opinion.
- Lyx
[a href="index.php?act=findpost&pid=284392"][{POST_SNAPBACK}][/a]

I'm not fighting against an encoder (I'm too old for this childish behaviour - or at least I hope so), but against some frequent TOS#8 violation due to a bad interpretation of the real sense of 3.90.3 recommandation (which isn't based anymore on a valid argument anymore I'd say).

I've just remind that we can't answer by "3.90.3" to the question of the best lame version. It's maybe the good answer, but there are no proof. Some people can recommand 3.90.3 because of old (should I say VERY old?) tests done before, but not because it's the best.

Best version of LAME and best arguments...

Reply #14 – 2005-03-22 08:46:09

What does it achieve if you get something through which is factually right, but damaging on a large-scale in the long term and which causes total confusion with existing practices (alpha builds are called alpha for a reason)?

I have high respect for all the stuff you did in the past. But this time, do you just want to be correct, or do you want to improve something?

3.97beta/stable probably is not far away, why the timing to do it now, when you can achieve both - be correct and improve something - in the near future?

- Lyx

Best version of LAME and best arguments...

Reply #15 – 2005-03-22 08:54:16

Quote

What does it achieve if you get something through which is factually right, but damaging on a large-scale in the long term and which causes total confusion with existing practices (alpha builds are called alpha for a reason)?

Correct me if I'm wrong, but it seems that the existing practices is to use the encoder which offers the most interesting features (gapless, replaygain, speed)... or a possible better quality.
aoTuV beta 2 was used on this board rather than 1.01 stable
mppenc alpha is currently recommanded; "beta" are also usually recommanded.

Only exception is LAME. Yes, contrary to other encoder, older lame was "tested" and that's why it's recommanded. But nowadays, there's nobody to seriously test new releases, alpha or not. In other words, 3.90.3 will be recommanded for years...

Best version of LAME and best arguments...

Reply #16 – 2005-03-22 09:04:47

Bah, i feel like a bulky conventional in this role. It's MY part to play the social maverick, not the other way around </joke>

Quote

aoTuV beta 2 was used on this board rather than 1.01 stable
mppenc alpha is currently recommanded; "beta" are also usually recommanded.
[a href="index.php?act=findpost&pid=284404"][{POST_SNAPBACK}][/a]

While the above encoders use the same terms for development-circles, their dev-circles have different characteristics compared to lame. For example, has there ever been a stable version aoTuV? No, they're constant-beta.

Quote

Yes, contrary to other encoder, older lame was "tested" and that's why it's recommanded. But nowadays, there's nobody to seriously test new releases, alpha or not. In other words, 3.90.3 will be recommanded for years...

Every veteran on this board who has been following lame feels it in the air that 3.97 will be the showdown. Even moderators have mentioned this feeling. Together with the recent usage-statistics, it is highly probably that 3.97-stable WILL become the recommened version - officially or not. You just would need to wait a short while and be correct and right.

- Lyx

Best version of LAME and best arguments...

Reply #17 – 2005-03-22 09:14:21

I for one don't see a problem in saying that LAME 3.97 alpha is with all probability better than 3.90.3.

Everyone has the choice whether to stick with the old-'n'-tested or to take a little risk (so little it is hard to define) for the reward of all-around better results. Telling people that there is a choice certainly isn't a bad thing to do but really a must.

Best version of LAME and best arguments...

Reply #18 – 2005-03-22 09:14:28

And MPC?

Best version of LAME and best arguments...

Reply #19 – 2005-03-22 09:44:30

Quote

Every veteran on this board who has been following lame feels it in the air that 3.97 will be the showdown. Even moderators have mentioned this feeling. Together with the recent usage-statistics, it is highly probably that 3.97-stable WILL become the recommened version - officially or not.

Dev0 asked for additionnal tests. lotr did some other tests, but that's all. It was 10 days ago... In these conditions, we're still very far from recommanding something else than 3.90.3... Now alpha9 and alpha10 were released, but ABX tests= 0.

On the other side, most people are still recommanding 3.90.3 as parrots, but for unclear reasons they don't contribute to test latest alphas and rarely provide ABX results. At best, just very vague claims (like this one)

I just have the feeling that most people are talking about quality at high VBR bitrate without being able to test anything.
HA.org = 20.000 members
HA.org total posts per day = 200
HA.org ABX tests posted per month = 1

And that's what some people are calling a "scientific" board or an objective one. It looks more like Scientifics in a hammock than scientifics with headphones.

Sorry for being haughty, but I'm highly disappointed by the current quality of HA.org board. Developers are doing a wondeful job, but I can't say the same thing from most common members (fortunately, there are exceptions: bug reports, some ABX tests...). LAME 3.90 was tested when some HA members were more active than current one. Those days are now over.

Best version of LAME and best arguments...

Reply #20 – 2005-03-22 10:00:05

I think that recommanding an alpha version of Lame is a little problematic.
If we do not consider the theorical alpha/beta/release stability status, there is still a problem with the way we release Lame.

Beta versions and releases have a tag in the cvs, so they can be pulled without any problem.
Right now, we are using the cvs to our daily developements. This means that from one day to another one, the content can change. The problem is that there is no tag for alpha versions. So let's say that I update Lame to a11 the 25th of march. If the 26th I commit another change, it will still be in a11.
So someone willing to grab an alpha from the cvs has no easy way to know if it is the alpha in the same state as its released day, or if it was changed after.

Best version of LAME and best arguments...

Reply #21 – 2005-03-22 10:07:13

Well if I end up getting the sack this week, maybe I could learn how to spot artifacts and even do some ABX testing hehe [we'll see if i do get sacked or not eh]. I'll search for some info on how to do it, unless a couple of peeps have a few links to hand [nudge nudge wink wink].

On a serious note I appreciate the efforts of HA as a whole even if I don't post that much as I've learned so much compared to even a year ago. Provided I don't get the sack, I'd like to make a donation towards HA, so who do I speak to about this

To be honest I still use Lame 3.90.3 at aps only as it's the most tested and the other encoder I use is iTunes aac 128k.

Best version of LAME and best arguments...

Reply #22 – 2005-03-22 10:10:05

Quote

Provided I don't get the sack, I'd like to make a donation towards HA, so who do I speak to about this

[a href="index.php?act=findpost&pid=284440"][{POST_SNAPBACK}][/a]

God what an idiot I am, I've just seen the Make a Donation button, perhaps I'm having a blonde moment [yes I have blonde highlights before you ask hehe]

Doh

Best version of LAME and best arguments...

Reply #23 – 2005-03-22 10:13:49

Quote

On the other side, most people are still recommanding 3.90.3 as parrots, but for unclear reasons they don't contribute to test latest alphas and rarely provide ABX results. At best, just very vague claims (like this one)

Ah sense prevails! Bravo. I can only think of all the man hours wasted by people who install some modern software then feel compelled to search around and replace the Lame version with an older one on the word of those who have not run extensive tests of the later versions, very sad and very damaging to Lames development.

Best version of LAME and best arguments...

Reply #24 – 2005-03-22 10:50:19

Quote

I think that recommanding an alpha version of Lame is a little problematic.
[a href="index.php?act=findpost&pid=284430"][{POST_SNAPBACK}][/a]

I was maybe unclear, but I didn't ask for recommanding current LAME alphas. But there are facts: even on HA, alpha were (and are: see mpc for example) recommanded. Using an alpha never killed someone, and rarely ruined encodings. I also recall that the first 'amazingly tested' 3.90 was quickly replaced by 3.90.1 in order to correct a critical bug found one week later. And that 3.90.2 had severe issues with some samples which were not revealed by the 'intensive' tests before.
Consequently, if alpha can't be recommanded for rational reasons, the same kind of problems could occur with stable release, even 'recommanded one'. But apparently, most people ignore this, and are just happy to repead everywhere that 3.90.3 was highly tested, without questionning this fact (what kind of tests? tested against what? 3.90.3 was never tested - it was 3.90 - but is nevertheless recommanded; tested by how many people?...).

Notice