Sorry I don't have the time for an adequate test, but I'm leaving for holidays tomorrow.
I just went quickly through my usual samples using -V2, and in an overall view I'd say there isn't a big change compared to b3. Some samples are even bit-identical.
Exception is eig. To my ears eig behavior is further improved, and if there weren't this little spot at ~ sec. 3 the result to me was pretty perfect. Very good result for such an mp3 hard sample!

Indeed
eig got better. You can still hear smearing, especially on these clicks at second 2-3 but it is far from being an annoying artefact at -V2. 3.97 is horrible in comparison.
At least V2 gets really solid with these newer builds even if it needs more bits sometimes.
Since other kind of samples don´t change in bitrate i hope lame is getting better in handling such demanding parts by really detecting it.

Big compliment to the devs!

What did you find from these?

#Try ABC/HR for direct comparison next time. You always suffer from placebo effect unless you use ABC/HR.

lol! my 3.98b3 is 420.934k and b4 is 421.040k
and the big click at second 2-3 with b3 is not worth an abx.

Maybe you have the first compile released on 2007-05-22. AFAIK there were 3 version of beta3 on Rareware: 2007-05-22, 2007-05-25 and 2007-06-06. The last two compiles have completely identical output with beta4 and the first one could have a bug. history.html metioned every improvement version over verson and there are no improvement since beta2 regarding quality.

The difference you two guys heard are not improvement of beta4.

My b3 is from 22-05 and is different to b4. b4 to b2 is much different again and much improved handling this eig sample, even if there is no entry in the history.
Did you try to listen and qualify?

Edit: before more confusion arises. My b2 is from 16-05

I don't have beta2 and you don't have b3 (2007-05-25 and 2007-06-06). That makes us confused.

All the thing I can tell is the fix on eig was included in beta3 of 2007-05-25.

My b3 is from 2007-05-22 as well.
While I can understand robert (guess he doesn't want an inflation of version numbers) it's not a very good idea IMO to have different encoder versions under the same name.

the other beta3's on rarewares included patches not affecting quality, and were accepted into cvs beta4(pre).

and john33 manages the lame releases on rarewares, not robert(o).


later

I very much appreciate having access to a compiled binary but I find the rolling binaries a little annoying. Not their existence but calling them all the same beta when they have post-release patches. I'm not sure what the binary naming convention should be but minimally it should have a "beta x+" or something to indicate it is not compiled from the officially released source code.

Actually, I think he may have been refering to the lame-dev Robert (Hegemann), known as 'Robert' on here.

mmm yeah i dunno (not psychic)...

I often got files at more than 230 Kbps using V2. How comes?

LAME assumes that more bitrate is needed to achieve the quality defined by V2.

I know but it didn't happen with previous version! It was more about 190Kbps.
I'm glad I didn't choose V0

Good

LAME 3.98 is forced to use stricted ISO mode now, thanks to the buggy FhG mp3 decoder that ships with Windows and this also reduces the bits that is stored on the bit reservoir so LAME will need to use more high bit rate frames now to get more bits since the reservoir is smaller then on 3.97.

I failed to ABX the trumpet killer sample at -V 2 which is very easy to ABX on 3.97 at -V 2 --vbr-new, seems like they have made some major improvements on the new VBR mode.

The new tagging features are really nice you can now add custom genres and album art now without using a tag editor.

Does anyone know what the tagging --tv switch does?

I'm pretty sure that's not the reason as the size of the bit reservoir gets only problematic for some decoders if the total bitrate of a frame exceeds 320 kbps, which isn't a big problem in the V2 range.

More probably it's the recent tuning that was more about quality than quantity, which is in principle a good thing. But IMO the average bitrate range should be lowered at least to old 3.90.3 APS range, because people don't expect so high bitrates with V2.

from the lame SF patch tracker:
http://sourceforge.net/tracker/index.php?f...amp;atid=300290




later

Thanks for the information on that switch, i had a quick play with it and it is very usefull for Various Artist albums and custom tags and easy to use when find a list a id3 frames.

I used the --tv and ripped a VA soundtrack on EAC with this command:

-V 2 --add-id3v2 --pad-id3v2 --ignore-tag-errors --ta "%a" --tt "%t" --tg "%m" --tl "%g" --ty "%y" --tn "%n/%x" --tv TCMP=1 --tv TPE2="Various Artists" %s %d

This tags mp3s with the BAND tag and iTunes compile tag. WMP sorteted the soundtrack as a VA album and iTunes sorted it out as a compilation out of the bog without using a different id3 tagger.

But I don't know if foobar2000 uses a id frame for ALBUM ARTIST though.

I played around with this EIG and its like Florida seq sample from JohnV. 3.97 -V2 is heard outside abx at 2-3 secs. A bigger problem is @ 3~4 secs that is almost masked by the keys. -V0 and high. bitrate ABR not very noticeable outside abx. Anyway, 3.98 -V2 is better, ABR same as 3.97.. Still abxing 3.98 @ 320k was possible without much effort and I was under the impression that 3.98 was worse than 3.97 .. So going above -V2 [ABR 250~280k] has some uses, but going to 320k is useless as there are cases like these that are easily abxable @320k.

What can be done really about this? According to Dibrom and JohnV : Nothing without some major overhaul to the format.

I found something much worse than EIG, easily heard @320k and vorbis also suffers preecho upto Q7. @ Q8 the bitrate was 313k and I failed to abx but had the impression it wasn't 100%. Even at Q5 it sounds better than mp3 @ 320k.

Fortunately, it is 100 % artificial - not on a music CD.

link:

http://www.hydrogenaudio.org/forums/index....showtopic=56186

so -V2 now uses -q0 as default?

Does anyone here happen to know when beta 5 is going to be released?

I guess when it's ready...

There should be something posted by the LAME devs "when it is released"

You should also check http://lame.bakerweb.biz/ for automated bleeding edge compiles from the CVS

But then again, I guess more testing should be done on the current BETA! don't you think!

I love beta4 @ V2, but still prefer to use -V0 (I'd rather use -b320 but too big... probably paranoid or something)

PS: I can ABX V2 vs V0 as well as V0 vs b320 on some music thou, not all!

today or tomorrow, depending on where you live.

Full set of win32 binaries at Rarewares now.

Good news, Thanks!

Discussion regarding beta 5 should take place in this thread.