Set up a directory called 'takehiro' and place a copy of the command line version of cvs.exe in it. Then copy the following into a batch file (tak.bat, for example)



and type tak (return) at the command prompt. The branch will be copied into a sub-directory called 'lame'.

BTW, just hit the enter (return) key when prompted for the password.

Wow !!!

What can I say. Great work Takehiro (4.0), great work Gabriel (3.9X) !!!!

We all knew that Takehiro has doing lots of new stuff, but it's very nice to actually see it (thanks to you too, Jonh33)

I was just thinking... I really believe it's too early to even think about quality. We will probably have to wait a lot more to have the first usable version. But until then, we have lame 3.90.3 and the future 3.94, that looks great too!!!

So, guys, we must have some patience...

John, I'm a beginning programming student, and looking through the source I really have no idea what is what. Is there a table of contents included or something?

I doubt I can understand much of anything, but I'd still like to look at it.

thanks John33, i'm actually on linux, but i was able to adapt your instructions just fine ;-)

Do you know if this will be updated in this branch, or will another more "official" one be made soon? (bringing the total branches closer to that triple digit mark)

In the near future, Takehiro's work on v4 will stay in the same branch. At least it has been the case for several months now.

No TOC that I'm aware of. In simple terms, the 'dll' dir contains the frontend code for creating the dll using the libmp3lame encoding libs; the 'frontend' dir contains the frontend code for creating the exe using the libmp3lame encoding libs and the mpglib decoding libs.

The names of the files indicate pretty accurately what each is concerned with.

On aps 128 the standard qval is 5,why?

-q2 is not tuned yet?

I'd say it's more that nothing has been tweaked yet

Give them some time and remember that this release is many months if not a year off still

yeah i know...but lame4 appear to be so fast because -q5 is used...if u use -q2 the speed is the same as 3.93

Well,finally J-Stereo is standard on 192 kbps cbr encoding,it's a very good think,at least imho

What lame really needs now....

is a cloning vat for Takehiro.

Version 4.0a6 must be updated or something... It says: may 12 and creates slightly smaller files than the encoder initially posted.

You're quite correct! There was a library change and consequently, I updated it.

I encoded a tune (metal) with the "new" version, aps, and it turned out around 230 kbps or so. Then I added the -Y switch (my favorite!) and it came out as a puny 130 kbps file! That's a pretty hefty slaughter of bits. But then again, it's only a pre-alpha:)

It's just been updated again - dated 15 May. This is an ICL7.1 compile.

Same link as before: http://homepage.ntlworld.com/jfe1205/lame4.0a6.zip

Whoa! That's some funny math. Wouldn't it be Lame 11.84?

Encoding errors @128kbps

http://digilander.libero.it/simartig/test_128.mp3

it happens with aps cbr128,-b128,etc

EDIT:Only with Joint Stereo.With Stereo encoding there are no errors B) I'll upload the sample ASAP,then I'll make more tests.

So,here's the full stereo sample

http://digilander.libero.it/simartig/test_..._128_stereo.mp3

Are you talking about those glitches which resemble CD reading errors? I've noticed those with ABR encodes (when I tested it they happen at the same place the tracks).

Yes,you're right.Can u test if it happens using Stereo samples instead of Joint?

Killer Sample >:]

Try to encode this @ 128 kbps with Lame4:

http://digilander.libero.it/simartig/test_4_128js.rar

It clearly fail to encode with JS,with Stereo it's a bit better,but still fails.

Some encoded samples:

Lame 4 -b128 -q5

http://digilander.libero.it/simartig/Lame4_128js.mp3

Lame 4 -b128 -q5 -ms

http://digilander.libero.it/simartig/Lame4...4_128Stereo.mp3

Lame 394a14 -b128 -q2

http://digilander.libero.it/simartig/Lame394_128js.mp3

Can I just point back to the original post. This is a purely experimental, pre, pre alpha. I only posted the binary so people could get some idea of where LAME was going. Any kind of in depth testing at this stage is completely pointless.

I Get the point.

I know it's a pre alpha,but i think that some tests can help Takehiro to choose the way to go on.

Or I'm wrong?

You may be right, I don't know, but apparently Takehiro wasn't too pleased with me for posting it in the first place!!

Hehe ok,now i understand

I think Takehiro knows what needs to be done, and doesn't like the idea of having several boards members try to tell what to do.

Give him some credit, he is Japanese, not just some dumb American!

I don't thin anybody's trying to tell him what to do. People are just eager to inform him/HA when they think they found something to improve, that's all. If Takehiro gets pissed off, he should just ignore these posts, shouldn't he?

Besides, I have come to realize Americans are actually smarter than they appear to be..:-)

Does anyone have the originally posted builds of Lame 4.0a6 from this thread?
I stupidly downloaded the May 17th build over the older version, and it sounds rubbish in comparison! Everything has gone all warbly and splashy/mushy sounding (accompanied by a small bitrate drop) when using --preset standard -Y. The older build made great sounding files with tiny bitrates, and in super quick time, so I would really appreciate it if someone could link to the older build.
I dunno what happened with the newer build - one of the library changes must have screwed it up I guess; I also noticed that the "-q x" setting doesn't work properly in the new build, it only makes a difference at q0, and above q6 I think - q1 to q5 all give identical files, as does q6 to q9. Mind you, I have no idea whether the older build had this bug too.

Thanks in advance!

In my opinion audio quality testing on these pre pre alphas are not really that useful, because the codes are not even finished yet. Testing and tuning should be done at least when the program is fully functional.

dvdmeister: I think you should not use this pre alpha for your daily usage. Those weired screw-ups and bugs are expected. You should stick with release versions.

I'm not really testing this version of Lame - I fully understand that its going to have bugs, and I'm not using it to make archive files or anything like that. I'm just using it to make mp3s for my portable player. It just happens to make better quality vs bitrate files than the 3.9x versions (in my opinion) for this application - I haven't and won't be doing any proper testing on Lame 4 at this stage, I'm just using it 'cause I found it worked well for me. Well, the old build did anyway

Will there be periodic binaries released from the 4.0 branch or is this all we are gonna see for several monthes?

My bet is that Takehiro wouldn't like to see frequent binaries. He already got pretty annoyed with this one.

It actually went to alpha 7 on the 24th, but there is clearly a lot of experimental work being done and the quality is, consequently, all over the place and changes on every update. My own decision, therefore, is that to post binaries right now wouldn't be helpful to anyone and that numbers of experimental sub-standard binaries in circulation can only be damaging. However, if someone else wants to do so, that's their decision, but I have no desire to upset the guy that's doing all the valuable work (any more than I have already, that is)!!.

From a quality perspective, it would appear that I just got lucky with the first of the alpha 6's that I posted.

I agree with this...but a changelog between the alphas may be very appreciated

At least we could see where the developement is going

the link is broken, where can i get lastesr lame 4 alpha6? thanx

What's the status on the latest LAME 4.0 (alpha10) posted on Mitiok's page??

http://mitiok.cjb.net/

It's got to be worth trying out since it's (kind of) officially out.

This may be a 'Lame' question ( ) but is there a dll version of Takehiro's build of Lame available? I'd like to check it out with my ripper (CDex) if possible. So far, the only version I've found is an executable.

In no way it is officially out:

WE DID NOT RELEASED LAME 4.0

What you can find on Mitiok's page is a compile from a cvs checkout that should be used for experimentation and testing only.

I a bit confused with the actual purpose of those binaries.
My question is how exactly L.A.M.E developers can benefit from user feedback and what they really expect from us to help them out.

I might be a bit unfamiliar with exactly how Mitiok is affiliated with the LAME dev's. Anyhow, it's there for anyone to d/l, so it would still be interesting to hear something about the progress.

AFAIK if developers are intrested sound quality of the encoder, then hearing other people opinion from quality is important.

EDIT: I just remembered local saying but it propably isn't suitable from here because it is related to stool.

The binaries do NOT serve a useful purpose at the moment. Takehiro is not really interested in any test results at this stage because it is far too early in the development. When he wants tests done, he'll ask for them.

By all means try them out of interest, but serious comments about the results are inappropriate at this time.

LAME4 is changing almost everyday. And I know it should and will change for a while, because I already notice there're so many problems.

This is because I do not need any test now.