Topic should be:Dibrom&John and others, how do you see the future of LAME?

So tell me. I wanna know.

I really have no idea other than I think that nobody else (the developers) really does either. I can't see any good coming from this fact. I don't think it's good for making a clear progression or for motivation to make futher improvements. As for the rest, I think you'll get a different answer from each of the developers, if at all.

You might try asking Alexander on the lame-dev. He's the "unofficial" (I guess there is no longer an official) release manager these days. Maybe he makes the decisions. Somehow though, I doubt many of the highly quality oriented features you asked about in other threads (two pass vbr, adaptive lowpass, etc, along with many others which could be thought of that you didn't mention) are on that list. Don't quote me on that though, Gabriel might know something that I don't again..

BTW, you might try this thread:

http://www.hydrogenaudio.org/forums/index....&f=15&t=4544&s=

Difficult question. It's no doubt difficult time for Lame now that lots of new code by Tak will be incorporated into official Lame in 3.94.
I know one thing.. The next release shouldn't be rushed. I also know that many people will always use the latest versions, so I want to help make sure, that it doesn't turn into complete quality disaster. The 3.93 release woke me up to think what really could happen if nobody does testing anymore.
One reason why Lame testing has recently diminished is --alt-presets. On the otherhand it's very good; you don't see newbies posting insane switch-sets anymore. On the otherhand everybody is/was expecting that --alt-presets just keep uptodate, and the floor kinda collapsed, when Dibrom announced he doesn't continue with Lame.

I personally just want to help so that 3.94 and further do not turn out to be quality disasters. If at the same time some people can find even better working switches which can be incorporated into --presets, the better. But the danger is that we may start to see the "newbie wild lines"=insane switch sets, if we do this thing publicly. Nobody wants to go back to the situation where people without too much knowledge how Lame works, posted hundreds of "best" Lame commandlines.

Do you mean is time to switch to Vorbis ?

(i really hope so)

These are individual choices. Lame developers are not switching to Vorbis, at least according to my knowledge. HA already has complete Vorbis forums, so there's no "switch".
If you are asking whether Dibrom is gonna start tweaking Vorbis, I have no idea, but I think HA's "General - (MP3)" will remain as the most visited codec-forum for a long time to come..

@Patchworks

No ogg thread here!? Was there any OGG question?

Wombat

what i wonder is:

- what is supossed to be a good quality testing practice? (is it important to get as many ppl to the listening tests or few (specialists with golden ears) will do fine?
- is there a way to automate such procedure to some point?

(<sci-fi> like via some automated page and/or p2p system (combination of both) for transfering samples/new alpha lame-versions and results/comments from and back to the main database </sci-fi>)

smok3:

Yes, there is a possibility for a system like this. Olcios proposed something at the very least (check out the thread on the testing framework). The problem is that this type of system is only as useful as it is maintained by someone on the project.

http://www.hydrogenaudio.org/forums/index....=ST&f=16&t=4336

I missed that one, it is exactly what i had in mind, only a p2p/webpage-database would prolly be a better tech solution on the long run as to avoid bandwidth issues.


That means there is no real wish for such type of project?

I don't know, you'd have to ask the LAME devs. If I were to venture a guess as to whether or not it would be maintained well, I'd point to general state of LAME right now as an example of what would probably be reflected in another venture related to that, but one requiring significantly more involvement on a day to day basis than what is currently required even.

In short, I don't think it would work. At the very least, I'm fairly certain it wouldn't work without either:

1) someone new getting involved (like JohnV) and putting forth a tremedous effort against many different antagonists
2) a complete change in the way LAME is currently managed. The developers would have to take on much more responsibility and would need to be accountable to such a system for it to really work. I believe that at the very least, Takehiro has said that he is not really wanting to do this. I don't know about the other developers, but Gabriel has said he only has a few hours a week for LAME (which probably isn't enough). As for the other developers, they seem to be involved much more in a much more erratic fashion, so you're guess is as good as mine. Oh and Naoki is no longer involved in the LAME project (or anything else AFAIK).

Or possibly some other drastic changes which I can't think of atm.

<guesswork> i do wonder what the reason is, i mean is there no need for such project or other reasons... </guesswork>

speaking as a complete outsider to lame-dev i see the situation something like: lame-devs on one side making their best efforts (knowing that there is no real competition in the mp3 zone) and the user on the 2nd side, from the users point the encoder is only a 'black-box' with certain output.

There is a big gap and probably the only way to implement such testing framework would be a 3rd person/community capable of talking to both sides and also that way the framework would be usefull for any other encoder out there that way.

<sci-fi> what about certain xchat/bitchx/mirc <-> eggdrop(s) <-> main database - web page solution, certain scripts could be writen so everyone running his own eggy could participate (sort of a half-p2p solution) </sci-fi>

The problem is that I don't think we're seeing the "best effort" from the developers. This may seem like a rude thing to say, but I honestly don't believe that most of the people who work on LAME could say they are giving it their best, certainly not if they are being honest. Furthermore, if people have the time and energy that it takes to argue about trivial matters (3.90.2 not being "official", 3.93 not being buggy), then they also have time to be coding and making improvements, or to use that energy to create a better development framework. It appears to me (and I've seen it said this way specifically) that most of the LAME developers view LAME as simply a pet project or hobby.. something to do when they are bored and have a little bit of free time, not something that is a real priority. This is fine, but it becomes a problem when the program has such a huge userbase. I discussed this before, and I think that it basically boils down to a matter of responsibility. If the developers do not want the responsibility that a project like LAME requires, then they should do something to remedy the situation. I believe that they owe at least that much to their users. Releases like 3.93 are unacceptable.



I believe that I was that 3rd person for awhile. The problem is that for this situation to work, both parties have to participate equally. There was no shortage of users willing to provide feedback to the developers to improve quality, but there was little interest on the developer side in using that information and constantly providing improvements (again, we're seeing some now.. but a year after the fact, a time when many have lost interest in waiting). Furthermore, I feel that there is a bit of disdain for the common user from some of the developers. I saw some of this first hand when Alexander (the "unofficial" release manager) came to this board and implied that we were all basically a bunch of whiny windows users begging for more features, when in fact we (or at least I) were concerned that 3.93 would be released without proper testing and with quality issues. Lo and behold....

I feel that the biggest problem of all though is that I seem to be unable to reason with the developers on any of the issues which have come up. In most of the proposals I've made, I believe that I've presented a well reasoned and valid argument. In return, I see a lot of "missing the point" and simply completely misinterpreting or not understanding (or ignoring) the problems at hand. It seems that there is some sort of poor communication along the line. A good example of this recently is Gabriel's comment that this whole situation is basically about a "different frontend" that users here want. I don't even know where this notion comes from because it's so far from the critical point of the matter that if one were to interpret the argument this way, there could really be no fundamental understanding (hence it's no surprise that we have such a big problem here). Another example would be the one I pointed out above, with Alexander confusing the fact that we wanted 3.93 to be "quality oriented" with meaning that we wanted it to be "feature oriented". Ironically enough, the concerns of this board were ignored, and the "stable" release that Alexander spent so much time arguing about (which was basically the same thing we all wanted, we just wanted to approach the matter differently, or dare I say more properly) never actually came to be.

For more information on this, I suggest going back and reading this thread:

http://www.hydrogenaudio.org/forums/index....=ST&f=15&t=3836

Particularly ALeidinger's (Alexander, the "unofficial" release manager) comments. As you can see, a lot of effort was put into arguing as to why Takehiro's code shouldn't be used for 3.93, and why the 3.93 (the buggy one that was recently released) indeed had no quality problems other than the fast presets. Alexander "knew" this from reading the code, from the fact that developers perfomed "listening tests" (questionable in itself), etc, etc, etc. And then the bit about Windows users. Oh, and I was accused of seeding FUD . In the end, 3.94 (or Takehiro's code) was tested much more than the released 3.93, which ended up being buggy despite Alexander's vigorous assertion to the opposite. Oh well. After reading that at least, it might become more clear, even from only this single matter, how disheveled the development process is and why I am ultimately fed up with it. I also feel that it is getting worse, especially as the confusion continues over what will be in 3.94 or 4.00 (or whatever it will be) and what is what in the different CVS branches.



The technical implementation is trivial really. It's who is running it, and how, that is important.

<guesswork>
lame goals are:
- better low bitrates (IS and stuff)
- to tweak and add new presets (after new psy model?) in the entire range to get to the fullscale quality-based encoder similar to vorbis
or what?
</guesswork>


it does look like a hard task to implement considering all the known and half-known facts to me. I had HA community in mind of course.



that is a no-win situation...


i did read that thread some time ago and i must say that it looks kinda personal (and funny at the same time), so without knowing the exact background or being involved into lame-dev the end user (me included) can't really understand whats going on, or whats up with (or where they came from) 'front-ends' and 'feature oriented' stuff like that, although it is pretty clear that average end-user is not really important (for some lame-devs).

Dibrom; if you were a lame-dev and weren't happy with the way it was going, why not pull out and go solo? I've read that you have quit the lame developement scene, but what now? Surely you guys can get a team together. As you'll notice, i'm new. But, I see 3 people that could start the team out here, John, Takehiro and you, yourself, dibrom. Power to the people and all that jazz.

Peace.

To Dibrom:

I've heard and used the --alt-presets and so I'm convinced that you really care about quality (i.e. sound quality). I trust your judgement in this matter and I guess until you publicly announce that e.g. LAME 3.96 or whatever sounds better than 3.90.2 (or there's such a huge difference that even I can hear it without hours of ABX-ing), I will just stick to using 3.90.2. And I don't think I'm the only one who thinks that way...

So my question is: are you going to test future LAME releases or provide any sort of comment on them?

If Dibrom quits and you don't trust anybody else's listening, then it seems you are stucked with 3.90.2 for the rest of your mp3-usage...

Well.. update on the situation. Now it seems Takehiro is no longer interested in listening test data for LAME, because HA is "rude". FIY, to those unaware, he was the only developer recently who was working on the core LAME psymodel issues.

In http://www.digital-inn.de/showthread.php?t...?threadid=16971 , Jean-Luc posted :


I couldn't help answering

This doesn't look good for open-source. If LAME were a commercial project I doubt this impasse would persist.

Wow. The drama continues. If personal differences keep popping up and progress on LAME slows to a crawl, I'll seriously have to consider switching over to Ogg Vorbis or MPC as my preferred format. The only reason I haven't switched over yet is a lack of hardware support. The current LAME situation, as it stands, is not good for MP3, period.

Well, yes... exactly that's what I'm afraid of... :'(
But who knows the future? I'm not saying I won't ever trust anybody else's listening it's just currently I trust Dibrom's the most.

Maybe it's time to switch to Vorbis, or MPC once it has hardware support... B)

Well I'm stuck with MP3 for the time being considering the 3 different portables I use daily (HD, CD and CF card). And 'time being' is probably going to be more than a year and a half, since one new (other codec) portable couldn't possibly replace the niches each of these three occupy in my life - not in the forseeable future anyway.

I feel sad about this downturn. It's all about communication, isn't it? And communication is what this medium is supposed to do best.

Whoever said that this confusion wouldn't be happening if it were a commercial enterprise was probably right. A successfull enterprise takes account of the public and how they use a product. The old reality test.

Presets are a large part of what makes Lame popular. Seems important.

My name is Alan and I'm an XP user. Nex

(After sinclair, atari 8 bit, st gem, mpe, amigados, unix, linux, beos and even some apple)

In my opinion this whole debate and the accompanying emotions should be toned down a bit and taken into context. It's not like this is 1996 and all we'll be able to use for the next 10 years is mp3 . . . we have at least one and possibly two very good, open-source, patent-free (ogg vorbis & mpc sv8) codecs up and coming. Our emotions are better spent supporting those efforts, in my opinion, than rewriting Gone With The Wind with hours upon hours of debate and malodrama that's ultimately all for naught, considering that mp3's death-knell is presently sounding.

I was thinking the same thing, Mith (but didn't really want to say anything ).

Perhaps the open source projects with a "Great Leader" at the helm (like Linux, with Linus Torvalds) are the only ones that really succeed... either that, or single-developer projects. Seems like so many others just languish, or are subject to issues like this one...

This really is bad news. Seems like lame development has just virtually halted in this last week or so. I am particually concerned with Takehiro's comment about HA being "rude". For a start, i have never really read any "rude" comments concerning developent (or anything else come to think of it), and secondly, HA is a great source for information about high quality audio compression, where users/members have found it when looking for high quality audio tools/information, thus are concerned with quality. We all try and pitch in with helping to develop lame via listning tests, abx results and reporting bugs. If this is being "rude" then I fail to see how lame will continue to develop.

I'm not a developer or an expert in audio compression, but as an end user all i require of lame is to be of the highest audio quality and (if possible) the quickest. Speed isnt as important as quality though, and it really DOES NOT matter about features or the frontend! Therefore i see no alternative but to perminantly use MPC, even though there is lack of hardware support, as development and planning are currently ALOT better than lame's, and the quality is constantly improving!

Here are the directions that i would like lame to take (As an average end user):

1. Improvement in quality with every release, even if releases are rare/infrequent
2. Simple set of command line presets, similar to MPC and Vorbis quality settings (q0-q9 etc) and alias like Internet/Streaming, Portable, CD, Standard, Extreme and Braindead.
3. Releases are to be fully tested by Golden Ears before being publicly released, then released as Betas/Alphas for extra quality tests and bug reports.
4. Lame Developers to get on with each other, address problems promptly and thourghly, and to be concerned with only improving quality rather than features. Leave frontends to other developers.

Development is something that i know very little of, but reading comments posted by developers, i can clearly see that there is definatly turmoil in the dev camp. It seems like some developers want speed and quick releases, and others want stable and high quality sounding releases. In this case, it is and always will be a bad way to work as devs won't be able to work TOGETHER! It also seems that most of the developers dont care what the end user wants from their project, this is a very bad way to develop a project of this type. What lame needs right now is a main project leader (such as dibrom) to lead lame development, to set main targets and goals for other devs to achieve just to add direction to lame. This leader would be able to get feedback from HA users regarging what they want from lame, so that he can point the lame project in the right direction. Without this, i'm afraid that lame will no longer be the king of mp3 compression when bad releases are constantly released (like 3.93) as users will be put off using it. I among with meny others trust dibroms points of view and will most probably stick with 3.90.2 or 3.92 until a release supersides the quality of it. But the way things are going right now, i doubt it will happen anytime soon...

That's what we all seem to want - it could be so simple! But they obviously don't give a damn about it.

BTW, "frontend" refers to the LAME command line interface in this case, not graphical frontends.


I don't at all understand this bit about Takehiro leaving. He was never flamed or treated unfriendly - quite the opposite really, Dibrom thanked and credited him for his work wherever he could. I find this type of behaviour plain childish and weak.

Dibrom, did he give you any specific examples of the "rudeness" that disturbed him?

No. Too be honest though, I'd rather not talk about it anymore. I think that the discussion has really been pretty much exhausted. I said as much as I could, and it seems that it made little difference. I'm ready to move on.

Things are going from annyoing to worse to very bad.

Remember the happy times just a year ago? Dibrom's brilliant presets being fine-tuned and released with the 3.90? Was this the beginning of the end of LAME?

I guess the best thing is to let Takehiro do some more work and then try whether there is sufficient progress to look into tunning again?

BTW: I've been happy with LAME 3.9x and APS for nearly a year now so I haven't been vocal at all.

from lame-dev:


I promised to post listening reports to lame-dev in the future and he continues visiting HA's irc-channel.
He's also working on 3.94alpha5 based on the little testing I made.

I understand this because his time is limited, he only wants to get the exact precise info, and browsing HA and reading about how badly Lame is organized etc. (he only wants to code anyway) is taking time away from him, and isn't exactly any fun to read. Since Tak is the main coder atm, I wish other devs would try to improve the organizing of the project.

I'm not an expert on open source projects but being someone who has benifitted from all your work I just want to say thank you!. I'll continue to use MP3 for my portible needs until new hardware support comes to market for the other newer codecs. With all the different companies profiting from the MP3 format it baffles me why they wouldn't take a more supportive roll in the development of the codec. Again thanks to all who have worked on the project. B)

I can understand how takehiro feels... everyone is bashing the development process, even me! but the only reason we are doing is that one of the best developers of lame has quit which is proof that the development process is not working. It is a big downward step in the eyes of meny end users and obviously is going to be discussed alot on this forum. I hope takehiro will except our test results to further develop lames quality as it would be a shame and unproductive not to use such a resource that Hydrogen Audio has to offer.

@Volcano: I was referring to the same thing. It doesnt matter what lame looks like as most people use 3rd party frontends anyway. The CLI doesn't need to be changed as all it needs to do is provide functionality which can then be taken care of by 3rd Party GUI's which it already does...

isn't the whole frontend thing about removing redundant and confusing commandline switches? that was my impression.

My only problem with LAME is that the massive amounts of presets and switches go thoroughly undocumented. Even when someone asks what switches and presets are there?, no one seems to present a real answer, but go read and find out. In my simple opinion a list should be compiled of switches and presets and if one wants to know what they do then they can go and research it out, but a simple "cheatsheet" type organization of the options available really would be nice... I see no point in adding more and more options when 90% percent of people would never be able to find out what they were anyway.

LAME needs a documentation like vorbis ogg.

Leadership is getting someone to do what they don't want to do, to achieve what they want to achieve.

- Tom Landry

No one ever read the doc here?
For those who missed it, there is a part about the available presets in the HTML doc provided with Lame.

Well i think many people want to get the full documentation and a faq, just like vorbis ogg does:
http://www.vorbis.com/faq.psp

My symphaties for the developers, but I don't think that anyone has given them critique at a personal level. It's issues fighting, not people.

It may not be nice to read critique about work of you own. But of course, a coder wants to code. And I know this from business life, don't let coders mess with anything else than coding. After all one has spent a lot of effot and hours on it. But the critique shows that something is wrong. I remember when Dibrom was tuning his presets people supported him with test results and encouraging words. --APS and others were finalized and nobody has been nagging about them, people have been 100% satisfied. So why bring all this good down? It is quite predictable that community like this will react strongly if something they are used to is suddenly about to be diminished. And one more thing, this kind of "I will not play with you no more" attitude which is shown here and there is childish and non-respectful.

Sorry if I was rude - I hope I didn't cause any permanent damage.

The only good news is that 3.94a with JohnV:s commanline is almost as good as 3.92 --aps

Sorry to revive, but having just joined and discovered this forum, I must just say that things like this seem to happen when a project isn't a commercial project. When there's no motivation (and in today's society, money is a motivating factor!) there's no direction, or very little. Look, I started encoding using Xing and AudioCat, and eventually found Lame. At the time, I felt nothing could touch Xing (I didn't have a great setup and I understood Mp3's weren't completely perfect copies, plus I was using VBR ~160kbps when everyone was using 128) until I used Lame. Then I got more space, a better system, and more demands. Having read a lot of links, I believe there's not much more we can push out of Mp3 and Lame in turn. Probably speed, but then there's the speed vs quality issue. If you're not happy with Lame 3.90.2/3 output, then it's time to look elsewhere. I've seen the "Lame 4 alpha" speeds and listened to the quality. Look, speed's all you really can push now (IMO), but if you're using Lame in the first place, speed wasn't an issue. Pushing more commercial companies to use Lame as an encoder would be a way to get more development done, but hey?, what more can we ask for from a streaming format that has some limitations. Looks like a typical Betamax vs VHS. I don't think the developers are going to spend much time on a project that's pretty much done it's job. It got us to Lame 3.90x, and that's pretty cool for my purposes. (I don't mind my PC working in the wee hours of morn encoding B) ) My thanks to the guys who got us this far and saved us from Xing!!

I would like to see Dibrom take the LAME code and split it off. Do his own thing. Change the name. Make it better. If he had the time and the motivation.

3.90.3 is already pretty darn good. Gapless support is probably the only thing I wish were there, and I don't know if that would ever be possible anyway. APS is already transparent for me, and the only improvement would be to get the bitrate down.

I wish LAME were still improving, but I think we've reached the point where it's already good enough for portables, and for anything else we should probably be going to a different format anyway.

I can see things here that aren't meant as disrespectful, but to someone who has put hours of free time into their hobby only to recieve criticsm like this (granted, amongst a lot of heart-felt praise) it is going to come across as arrogant and downright insulting.

My opinion shouldn't be taken as worth much because I know so little about the matter. I however feel it is unfair to say the LAME devs aren't working hard enough, or 'could do better'. If their commitment doesn't satifsfy you, get involved yourself (for the first time, or again in Dibrom's case) These people are coding for your pleasure, not for their own.


I understand your point here, but maybe this is why we are seeing devs quit, or have become dispassionate with the idea of a unified testing method - they want to enjoy a bit of coding in their personal time, not have the weight of a million users on their shoulders.

Raising an opinion I have voiced before, I would love to see the devs move off to new projects. This has given me another reason why. FAAC is a fledgling encoder in need of the kind of work they can give. It doesn't come with the expectation of the world - and with it, the disappointment and dissatisfaction from users who think they're not getting enough.

Vorbis is another promising project crying out for help, 1 man split between 100 different projects, and so nothing ends up getting done in good time. I see helping MP3 as a waste of good talent, the format is 99% squeezed dry by the LAME codec, whereas AAC/OGG are what, 50% of their potential.

I just wish some of the coders felt the same, we could have progress in new technologies that will really forward the world of audio compression.

I see vorbis as a project thats forever going to be wedged between more compatable and widely supported formats. No offense to people who like vorbis but its worthless to me if it never gets any portable support. For that reason squeezing more from MP3 makes sense- it will be put to good use by lots of people.

Anyway with LAME4 apparently in the works this whole thread is dated.

Not actually... what's left to squeeze out of Lame? Is a 1% increase in encoding efficiency worth all that effort and time, when promising new codec's that could use that development time could improve their functionality dramatically for the benefit of the masses in the future.

Oh, and this is the future of Lame in general and I don't think Lame 4 is going change the question. Its supporting a format that's getting a lot of flac from other format developers thanks to the commercial and technical limitations of Mp3 and this places Lame in the hotseat. Do we continue to develop something already optimised or use the time to get something better?

why arent the lame developers interested in improving the opensource aac encoder faac?

Remind how many commercial devices support MP3 vs. AAC/ogg/mpc

I'm not sure, I've asked this before but don't remember an answer (i'm forgetful!)

Gabriel?

How many Lame developpers do you think there are?
How much time per week do you think they have?


If I would move to something else, it would probably be aac. But I do not want to let Lame as-is, knowing that some things can still be improved.

My guesses would be 4 and 1, which explains why LAME moves slowly

Thanks for answering, it's something I've wondered for a while