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 3.90.3 will never be retired (Read 12044 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lame 3.90.3 will never be retired

Let me preface what I'm saying here is that this is by no means an attack on the fine 3.90.3 recommended encoder.  I think it does a wonderful job. But I do however want to throw some questions.

After the last round of lame listening tests and 3.90.3 stayed the recommended version, I started wondering what is it gonna take to get the "best" version updated. I'm not being impractical here but I really hope it wont stay locked at this version forever say 10 years from now!  The main thing that sparked my interest in this matter is post by rjamorim. Baisclly, he brings a very valid point about the way in which the encoders are being tested. The original was tested how well it worked against problem samples while the newer one is being tested directly against another encoder. My main question I guess is this an accurate/fair comparison?  Moreover, can 3.96 hold out well against problem samples itself?

Personally, I'm mostly interested in APS. If a newer encoder can do as good of a job as the older one at a lower bitrate (and/or faster encoding) I think it should be recommended. I think though 3.96 failed on those tests at aps  but outdid the older 3.90 at low bitrates.  I'm really holding alot of hope on lame 4.0 and just dying for it to be released as a stable version so it can be compared to the older encoder.

EDIT: typos

Lame 3.90.3 will never be retired

Reply #1 –
Quote
If a newer encoder can do as good of a job as the older one at a lower bitrate (and/or faster encoding) I think it should be recommended. I think though 3.96 failed on those tests at aps  but outdid the older 3.90 at low bitrates.  I'm really holding alot of hope on lame 4.0 and just dying for it to be released as a stable version so it can be compared to the older encoder.

EDIT: typos
[a href="index.php?act=findpost&pid=262704"][{POST_SNAPBACK}][/a]

IIRC the handful of tests that were done showed 3.96.1 to be roughly on par with 3.90.3 in terms of quality at aps.  The administration of this board did not feel that these tests were enough to prove that 3.96.1, in terms of aps quality, is as good or better overall (on average) than 3.90.3, and I agree.  There were not enough tests for statistical significance over all types of music.  Thus 3.90.3 is still the recommended encoder.  However if you feel that the evidence that 3.96.1 and 3.90.3 are more or less the same quality while using aps is good enough for you, personally I do, then you would be better off using 3.96.1 over 3.90.3 due to the decreased filesize and the faster encoding, you do not have to use the recommended encoder.  As far as the recommended encoder being 3.90.3 forever, I do not believe this will be the case.  LAME 4 will probably have enough benefits in terms of speed and features that, I think, the members of this board could be compelled once again to build presets for LAME 4 the same way they did for 3.90.  Even if this is not the case I expect that LAME development will come to a point where the quality will be better then 3.90.3 to a large enough degree that it will not take so much rigorous testing to prove its superiority.

edit: clarification
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame 3.90.3 will never be retired

Reply #2 –
It was the Forum's management decision to keep LAME 3.90.3 as the recommended version...unfortunately

Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his  work to be burried...they'll most likely only change their minds if every single sample has been proven to be better no matter what setting...which is just impossible...
--alt-presets are there for a reason! These other switches DO NOT work better than it, trust me on this.
LAME + Joint Stereo doesn't destroy 'Stereo'

Lame 3.90.3 will never be retired

Reply #3 –
Quote
Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his  work to be burried...
can you by all means backup this wild claim Jojo?
Nothing but a Heartache - Since I found my Baby ;)

Lame 3.90.3 will never be retired

Reply #4 –
Quote
Quote
Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his  work to be burried...

can you by all means backup this wild claim Jojo?
[a href="index.php?act=findpost&pid=262853"][{POST_SNAPBACK}][/a]


I agree with you, although I think that it is why he employed the word "might". Sincerely, I'd find it quite disappointing if Dibrom reacted like this, because his good work is perpetuated on latest versions of Lame, as far as I know. And on 3.96.1, you can use his tweaks on all the VBR scale! Isn't it recognition for all the work?

Lame 3.90.3 will never be retired

Reply #5 –
Quote
I agree with you, although I think that it is why he employed the word "might". Sincerely, I'd find it quite disappointing if Dibrom reacted like this, because his good work is perpetuated on latest versions of Lame, as far as I know. And on 3.96.1, you can use his tweaks on all the VBR scale! Isn't it recognition for all the work?
[a href="index.php?act=findpost&pid=262856"][{POST_SNAPBACK}][/a]

The original DM presets were broken in subsequent versions of LAME.  The --preset and -V settings in 3.96.1 do not use the same tweaks as the --alt-presets from 3.90.x.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame 3.90.3 will never be retired

Reply #6 –
Quote
Quote
Quote
Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his  work to be burried...

can you by all means backup this wild claim Jojo?
I agree with you, although I think that it is why he employed the word "might".
well, in the light of general tone and style of his post, he could have just said "I think that they are biased etc".
in my view, 'some ppl might say' equals either 'on the one hand', followed by a contra, or it is a 'I think', while not having to need to explicitly say it is ones own opinoin. in that case here, I would strogly suggest the second possiblity. which would be nothing but FUD, if not backed up.
Nothing but a Heartache - Since I found my Baby ;)

Lame 3.90.3 will never be retired

Reply #7 –
Quote
The original DM presets were broken in subsequent versions of LAME.

How should we interpret this? Does this means that according to you the current version (3.96.1) has broken DM presets that cause its vbr encoding to be worst than 3.90.3?
You know that the 3.90.3 vbr presets have not been proven to be better than 3.96.1, don't you?

Quote
Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his work to be burried

I think that you are too much radical here. I think that it is more something that could be considered as "over-safetiness".

Lame 3.90.3 will never be retired

Reply #8 –
Quote
Some people might say that they are biased and looking for any reason just to keep LAME 3.90.3 since Dibrom was involved in some tweaking and probably doesn't want his  work to be burried...they'll most likely only change their minds if every single sample has been proven to be better no matter what setting...which is just impossible...[a href="index.php?act=findpost&pid=262834"][{POST_SNAPBACK}][/a]

That is complete nonsense. At the time the 3.90.3 vs. 3.96 test was conduced only tigre and myself were involved in it. (for the simple reason that we were the only HA.org moderators still interested in LAME development)
Dibrom hasn't made a single statement regarding the whole matter and we only acted according to the principles HA is build on:
The test was done in a completely unbiased fashion and didn't provide enough significance to justify replacing 3.90.3 as the recommended version (at least at --alt-preset standard and above).
One might argue that this was the wrong decision to make and that moving along with LAME development would have benefited both users and LAME developers, however it wasn't biased in any way.

In the past months I've tried cleaning up the recommendation threads to express the ambiguous situation of the matter.
"To understand me, you'll have to swallow a world." Or maybe your words.

Lame 3.90.3 will never be retired

Reply #9 –
Quote
The test was done in a completely unbiased fashion and didn't provide enough significance to justify replacing 3.90.3 as the recommended version (at least at --alt-preset standard and above).


That's a little silly if you ask me. As long as aps quality didn't regress on all but a few samples, why shouldn't it be recommended!? We all know 3.96.1 offers other significant advantages over 3.90.3; faster encoding, smaller file-sizes and better quality at lower bit-rates..

If aps is transparent, how could 3.96.1 possibly improve on that except smaller file-sizes (which it did) 

Lame 3.90.3 will never be retired

Reply #10 –
Quote
Quote
The original DM presets were broken in subsequent versions of LAME.

How should we interpret this? Does this means that according to you the current version (3.96.1) has broken DM presets that cause its vbr encoding to be worst than 3.90.3?
You know that the 3.90.3 vbr presets have not been proven to be better than 3.96.1, don't you?
[a href="index.php?act=findpost&pid=262898"][{POST_SNAPBACK}][/a]

Please read my first post in this thread .  I know this doesn't mean that the newer versions of LAME are worse, they very well could be better (I think at preset standard they are about the same).  I have already written far too many full explanations of how I precieve this; one is, as I said already, part of this thread.  I'm sorry if you feel that the above mentioned post could be misinturpreted.  I didn't want to, yet again, write out a full explanation; I wrote a specific reply to a specific post.

edit:  wow its amazing how with slightly altered punctuation my third sentence miraculously becomes legible
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame 3.90.3 will never be retired

Reply #11 –
Quote
The test was done in a completely unbiased fashion and didn't provide enough significance to justify replacing 3.90.3 as the recommended version (at least at --alt-preset standard and above).


dev0: thats exactly what I wanted to address. I think everybody is getting a little side stepped from what my acutal questions were. Thats one of them. WHat will it take to make 3.96 or whatever newer version replace 3.90.3 as the offical "recommended" ?

The orginal post I quoted initally had 2 schools of thought on this it seems:

1) test 3.96 directly agianst problem samples
ot
2) test 3.96 compared to 3.90

The later in the opinion of seem users here seems will and can never be meet. Nor is it realistic to expect any compile to "beat" another on evert possible sample.  What if 3.96 for the most part sounds great @APS on most of the problem samples?  Im not making these claims myself I dont know.

I am rather asking of those 2 testing methods which would be better for lame's recommended encoder? And why. I see pro and con'sfor both.

Lame 3.90.3 will never be retired

Reply #12 –
@ LiTEMaTTeR:  If you conduct several tests yourself and post the results, maybe you can kindle enough interest in testing 3.96.1 and it will subsequently become the recommended encoder.  I use 3.96.1 and if the fact that it is not recommended by HA bothered me enough to do what I have above suggested, that is the course of action I would take.  However it does not bother me very much at all, only very slightly do to my empathy for the LAME development team but I don't think they are losing any sleep over it; so I am certainly not about to.  You seem to be the most bothered by this ATM so perhaps you could start the testing, if so I might even be persuaded to follow  , no promises though.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame 3.90.3 will never be retired

Reply #13 –
Quote
@ LiTEMaTTeR:  If you conduct several tests yourself and post the results, maybe you can kindle enough interest in testing 3.96.1 and it will subsequently become the recommended encoder.  I use 3.96.1 and if the fact that it is not recommended by HA bothered me enough to do what I have above suggested, that is the course of action I would take.  However it does not bother me very much at all, only very slightly do to my empathy for the LAME development team but I don't think they are losing any sleep over it; so I am certainly not about to.  You seem to be the most bothered by this ATM so perhaps you could start the testing, if so I might even be persuaded to follow :) , no promises though.
[{POST_SNAPBACK}][/a]



"Bothered" isnt the right word at all!!! LOL!

I was just asking about it :) This is not so much for my *own* personal encoding desires. I'm looking at what is recommended for most people across the board i.e. the HA.org suggests. Which believe it or not does have some weight when you need to reference a site to a large audience (without me going into too much detail here).

Well there are [a href="http://www.ff123.net/samples.html]FF123's problem samples[/url] for APS to test. If 3.96 can handle them well (even if not *always* better than 3.90.3 itself) would it become the hailed champion encoder though? That is what Im' asking before even such a test begins.

3.96 vs problem samples directly OR 3.96 vs 3.90 (on prob samples)  <--- which method will the powers that be around here consider for the winning encoder version? Because the later might never happen from what I take it from rjamorim's posting in the "recommended LAME settings" thread and I fear he might be right. Maybe I have things completely wrong so somebody please educate me on the matter.

Lame 3.90.3 will never be retired

Reply #14 –
note: there are strong possibilities that with cbr/abr encodes lower than 160kbps 3.90.3 is still superior to 3.96.1. I think that this point is the strongest one against 3.96.1. (this might change in future versions)

Lame 3.90.3 will never be retired

Reply #15 –
All of this is really of little consequence.

There is no bias in the recommendation of 3.90.3 over 3.96.1, nor, indeed, is there much discernable difference between the outputs of both for the vast majority of normal samples. The majority of people are likely to be equally satisfied whichever encoder they use.

Credit should really be given to the devs, particularly Gabriel I think, that they have managed, in 3.96.1, to remove the need for the special tweaks that Dibrom had to use to achieve the level of quality attained in 3.90.3.

For anyone who has done any testing with the latest 4.0 alpha, I am sure they will agree with me that it can only be a matter of time before 4.0 becomes the encoder of choice both in terms of speed and quality. In the meantime, I think the choice of encoder is really a personal one with pros and cons for both versions.

Lame 3.90.3 will never be retired

Reply #16 –
I really don't understand what the racket is all about.

3.90.3 has undergone extremely strict QA and it can be said with confidence that encoding results with --alt-preset standard will be excellent. You can not make the same statement for version 3.96.1 with the same confidence. That's all. It does not imply 3.90.3 is of higher quality.

What testing has been done, suggests that the overall quality is extremely close. I haven't seen any samples that cause severe problems with 3.96.1 that are fine with 3.90.3. So I, having a brain of my own and the neccessary time to consider the facts, decided to use 3.96.1 for all my mp3 purposes and accept the risk of lower quality but embrace higher speed and smaller files.

Anybody else who wants others to make the decisions for him (and that is perfectly fine) will want the tried-and-tested solution which is still version 3.90.3.

@LiTEMaTTeR: If we want to replace 3.90.3 then we simply need to do more testing. Feel free to coordinate that. My idea would be to test one sample per week. All you need to do is select the samples and prepare them for use in abc/hr. I'm sure the people here would help you put together a web page hosted on the hydrogenaudio server and provide the space/traffic for the samples.

Lame 3.90.3 will never be retired

Reply #17 –
Well. Maybe this testing should be done for a future version of LAME, since what does it help to test 3.96.1 if possibly a new LAME is to appear soon. And if we tested a beta version, a future final version could benefit from these tests.

But the LAME developers should be asked and heard first.

EDIT: Typo

Lame 3.90.3 will never be retired

Reply #18 –
Quote
Well. Maybe this testing should be done for a future version of LAME, since what does it help to test 3.96.1 if possibly a new LAME is to appear soon.

I doubt that such a test would start tomorrow, but yes, it is likely that just now is perhaps not the right time for such a test, if it is really needed.
I'd prefer adventurous people to test the next alpha version (ie the next 3.97a not yet available)

Lame 3.90.3 will never be retired

Reply #19 –
Quote
Anybody else who wants others to make the decisions for him (and that is perfectly fine) will want the tried-and-tested solution which is still version 3.90.3.
[{POST_SNAPBACK}][/a]

I found this post quite interesting: [a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=30358&view=findpost&p=262843]http://www.hydrogenaudio.org/forums/index....ndpost&p=262843[/url]

it's about a pretty popular Sony Discman from 2002 that skipps when using --alt preset standard (LAME 3.90.3.) but doesn't when using LAME 3.96.1

things like that should be taken into account when making a decison for other people that don't want to bother to got through all the details...I haven't heart about any compatible issues in LAME 3.96.1 that didn't exist in LAME 3.90.3...
--alt-presets are there for a reason! These other switches DO NOT work better than it, trust me on this.
LAME + Joint Stereo doesn't destroy 'Stereo'

Lame 3.90.3 will never be retired

Reply #20 –
Quote
In the meantime, I think the choice of encoder is really a personal one with pros and cons for both versions.
[a href="index.php?act=findpost&pid=262982"][{POST_SNAPBACK}][/a]


This is all that really needs to be said about this issue. Too many people here are looking for the easy "recommended" way out without testing their codec choice for themselves.

Lame 3.90.3 will never be retired

Reply #21 –
Haven't seen any comments on testing 4.0 or 3.97alpha.  I think most HA Forum members are relatively conservative and that's why 3.90 has remained their choice.  Though hearing of regressions is troubling...
 

Lame 3.90.3 will never be retired

Reply #22 –
Quote
Quote
In the meantime, I think the choice of encoder is really a personal one with pros and cons for both versions.
[a href="index.php?act=findpost&pid=262982"][{POST_SNAPBACK}][/a]


This is all that really needs to be said about this issue. Too many people here are looking for the easy "recommended" way out without testing their codec choice for themselves.
[a href="index.php?act=findpost&pid=263054"][{POST_SNAPBACK}][/a]

At the levels of quality we're talking about, telling people to test for themselves is counterproductive and meaningless. You might as well tell them to test different power cables and see which one sounds better.

Lame 3.90.3 will never be retired

Reply #23 –
Quote
Haven't seen any comments on testing 4.0 or 3.97alpha.  I think most HA Forum members are relatively conservative and that's why 3.90 has remained their choice.  Though hearing of regressions is troubling...
[a href="index.php?act=findpost&pid=263086"][{POST_SNAPBACK}][/a]

It's the administrations choice, not the members.  The regressions were mainly from 3.90 --> 3.93/4.  From the relatively small amount of testing done on LAME 3.95/6 it seems that most of these regressions have been fixed.

edit:  I want to be clear about this; there still certainly are some regressions but there also have been some improvements too.
gentoo ~amd64 + layman | ncmpcpp/mpd | wavpack + vorbis + lame

Lame 3.90.3 will never be retired

Reply #24 –
Quote
Personally, I'm mostly interested in APS. If a newer encoder can do as good of a job as the older one at a lower bitrate (and/or faster encoding) I think it should be recommended. I think though 3.96 failed on those tests at aps  but outdid the older 3.90 at low bitrates.  I'm really holding alot of hope on lame 4.0 and just dying for it to be released as a stable version so it can be compared to the older encoder.

[a href="index.php?act=findpost&pid=262704"][{POST_SNAPBACK}][/a]


What prevents you from using 3.96.1 for your own music collection if these build's results serve you fine ? Just rip some of your favourite music and start to encode using both builds ... you can then easily judge whether your choice is 3.96.1 or 3.90.3. You don't really need a recommended encoder version to judge quality for yourself ...

BTW ... the idea behind a thoroughfully tested and thus recommended version is to give some general guidelines rather than telling users which version to use ... HA.org staff will, apart from some notorious zealots whose number did decrease over the years, never try to force anyone into using something particular (although there are some -knowledgeful- members that are definitely biased towards certain -capable- pieces of software).

Edit - Typo
The name was Plex The Ripper, not Jack The Ripper