hey guys,
someone knows which lame codec is the best one in 128 ABR? 3.97? 3.98a? or other?

thanks for replies

Best? I think you'd be hard-pressed to hear any audible differences between recent encoder versions at this bitrate. Try encoding a few samples with the different versions and ABX. They're all going to sound decent, perhaps even transparent.

If you're dying to split hairs, you might want to ask for halb27's input, he seems to do a lot of LAME testing with problem samples.

I did a test recently.
3.90.3abr132 was the best in my test among the Lame versions and settings I tested with a bitrate around 130 kbps. I tested Lame 3.97final -V5 --vbr-new, 3.98a11 -V5 and abr132, 3.90.3abr132, Helix -V55 -X2 and level's variant, as well as - on part of the samples - FhG surround CBR 128.

BTW I used abr132 in the test but found out later that abr130 is more adequate to match file size achieved by cbr 128 - well not too far away.

My test is in the Lame3.98a11 thread where you will also find an interesting test done by guruboolez using 3.97 and 3.98a11 -V5 --vbr-new.

what setting did you exactly use for your listening test?

just read guruboolez' listening test with 3.97 and 3.98a11 -V5 etc.

a bit depressing that 3.97 is better there but 3.98 better there.... no version that offers a compromise.... humm
which parameters did you use in abr for 3.90 halb?

Lame3.90.3 --alt-preset 132
Lame3.98a11 --alt-preset 132
Lame3.98a11 -V5
Lame3.97 -V5 --vbr-new
Helix -V55 -X2
Helix -V55 -X2 -SBT450 -TX0
FhG surround -br 128000

I think at ~130 kbps at least with mp3 we're in a world of compromise.

The challenginig thing is: we can't just take a listening test as is. Sure we want to use the 'best' encoder. But we have to answer the question: how bad is it really when an encoder behaves rather badly with a sample like eig, or Moon, or Tom's Diner, or viola de Gamba, or .... ? eig for instance provides a serious problem to any mp3 encoder, but if you don't listen to electronic stuff like that: does it matter much? In principle it matters as there is a chance that the problem in the machinery behind the problem can affect music you like. However it's a problem that is a lot less relevant than is the case when an encoder has a serious and not extrremely rare problem within a music genre you like or even is pretty commonplace.

So judging about suboptimal performance has a lot to do whether your favorite musical genres are seriously affected, and how your own perception is about the problem. From listening tests we learn a lot but we have to weigh the details and should listen for ourselves.

right.

but...
The Lame codec and the other lossy ones aren't just very good nowadays. Lame has become a great encoder since 6 years or more. This is crazy to say, but nowadays the way of lossy codecs is to perfection. Everybody can now feel free to use his/her favorite codec (When 3.90.3 / MPC was a recommendation that says it brings transparency, for cds it will always be transparency then).

Why is lame being developed anymore? because someone wants to create the perfect lossy codec, or am I wrong?

One day you can make transparent files @ 1 kbps ... no just a joke ... or is it possible?

I'll make some tests on my own with 3.90.3 3.97 and 3.98a11. Think that's the best solution for now. That doesn't mean I don't agree with the results of halb and guru, by the way, thank you all for the answers

so now I'll take some time for comparing tha big ones

It can be transparent for you @1kbps - especially when you're deaf.

It venture that it is already possible, you just have to have a huge bitdepth

Nonsense. You make it sound like they haven't evolved and improved.

2004: http://www.rjamorim.com/test/multiformat128/results.html

2006 http://www.listening-tests.info/mf-128-1/results.htm

I think you need to put things into perspective a little more and do those listening tests with music you listen to.

By just looking at the results it looks like there has been improvement (Lame goes up from 4.18 to 4.60).
In reality we don't know much about quality development. Both tests use a different test suite, and as long as a rather moderate number of people have contributed to the test results also depend on people.

One person can say 'Sample X has improved' by comparing two versions in direct comparison, but comparing different test samples tested by (partially) different persons by looking at the resulting quality number doesn't tell us much.

What we can say from both tests is that music is usually very enjoyable even at such a moderate bitrate. And that's the essential message I think.