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 3.96 released (Read 25816 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lame 3.96 released

Reply #25 –
Mp3 is a lossy encoding scheme, and it is not suprising to have some samples encoded better than others, and some encoded worst than others.

A regression on a few samples might not be problematic is more samples are improved.
So before talking about a "bug", please check more than 1 sample. If you find a lot of degraded samples and no one is able to find inproved samples, then there is a problem.

Lame 3.96 released

Reply #26 –
OK, who's going to start the LAME 3.96b2 vs LAME 3.96 final thread?
daefeatures.co.uk

Lame 3.96 released

Reply #27 –
Quote
OK, who's going to start the LAME 3.96b2 vs LAME 3.96 final thread?

from changelog:
Quote
*  Gabriel Bouvigne:
       o new quantization selection mode (used in ABR/CBR)
       o set sfscale for ABR/CBR up to 160kbps

So the changes done affects only ABR/CBR modes, no need to test VBR presets again.
Quote
Based on the tests done on 3.96b, cbr/abr modes have been improved since 3.96b2.

This is a statement based on listening test, 3.96 stable is better than 3.96b2 with almost all samples (ABR/CBR). Chirping/ringing problems i had noticed with previous versions are significantly reduced now.
WavPack 4.3 -mfx5
LAME 3.97 -V5 --vbr-new --athaa-sensitivity 1

Lame 3.96 released

Reply #28 –
I don't think testing 3.96b2 vs. 3.96 extensively is very helpful. The reference is still 3.90.3 as 3.96beta didn't outperform 3.90.3 anwhere (except maybe -V 5 vs. --alt-preset 128, but there haven't been enough results).

So starting a 3.90.3 vs. 3.96 final thread makes more sense. If bad regression samples are found, we could do additional tests 3.96 final vs. some beta version on developer request.

Thoughts?
Let's suppose that rain washes out a picnic. Who is feeling negative? The rain? Or YOU? What's causing the negative feeling? The rain or your reaction? - Anthony De Mello

Lame 3.96 released

Reply #29 –
Quote
,Apr 13 2004, 11:33 AM]
Quote
OK, who's going to start the LAME 3.96b2 vs LAME 3.96 final thread?

from changelog:
Quote
*  Gabriel Bouvigne:
       o new quantization selection mode (used in ABR/CBR)
       o set sfscale for ABR/CBR up to 160kbps

So the changes done affects only ABR/CBR modes, no need to test VBR presets again.

To be sure about this, it would be a good idea to perform some encode->decode->wave compare tests with VBR presets IMO. Additionally, most (alt)preset standard tests published in the 3.90.3 vs. 3.96beta thread were performed with 3.96beta1 - and minimum bitrate was set to 128kbps for --preset standard and higher VBR presets. (Most of) the few samples re-tested with 3.96beta2 didn't show much improvement, but that's not enough to draw conclusions IMO.

So tests would still be necessary for

VBR/ABR:
~256kbps: 3.90.3 --alt-preset extreme vs. 3.96 --preset extreme (not enough results)
~210kbps: 3.90.3 --alt-preset standard vs. 3.96 --preset standard (some verifications with 3.96 final needed)
~160kbps: 3.90.3 --alt-preset 160 vs. 3.96 --preset 160 vs. 3.96 -V 4 (not enough results)
~128kbps: 3.90.3 --alt-preset 128 vs. 3.96 --preset 128 vs. 3.96 -V 5 (not enough results)

CBR:
3.90.3 --alt-preset extreme vs. 3.96 --preset extreme
3.90.3 --alt-preset CBR <whatever> vs. 3.96 --preset CBR <whatever>

Quote
Quote
Based on the tests done on 3.96b, cbr/abr modes have been improved since 3.96b2.

This is a statement based on listening test, 3.96 stable is better than 3.96b2 with almost all samples (ABR/CBR). Chirping/ringing problems i had noticed with previous versions are significantly reduced now.

Sounds promising. I'd appreciate to see detailed results though , especially because of the "Recommended LAME version" question. - If there's interest I can start an "official" test thread...
Let's suppose that rain washes out a picnic. Who is feeling negative? The rain? Or YOU? What's causing the negative feeling? The rain or your reaction? - Anthony De Mello

Lame 3.96 released

Reply #30 –
Quote
If there's interest I can start an "official" test thread...

You previous set of threads (discussion, results, samples) was very helpfull, even with the low number of participants.

I would personnaly like very much to see a new set of ones.

Lame 3.96 released

Reply #31 –
Quote
So starting a 3.90.3 vs. 3.96 final thread makes more sense. If bad regression samples are found, we could do additional tests 3.96 final vs. some beta version on developer request.

Thoughts?

Before 3.96 stable release i've submitted Gabriel some results, my test is about --preset 128: http://xoomer.virgilio.it/fofobella/sfscale.zip
lamex9 is a test version gently provided by Gabriel, in the table the column named "lamex9 sfscale" performed very good according to my results. Lame 3.96 stable have similar behaviour up to 160 kbps.
For abr/cbr i would choose 3.96 stable over 3.90.3 overall, i haven't tested VBR presets but 3.90.3 seemed to be still better (and higher bitrate).

ABR/CBR modes are improved according to my results but testing with some more samples would be useful.
WavPack 4.3 -mfx5
LAME 3.97 -V5 --vbr-new --athaa-sensitivity 1
 

Lame 3.96 released

Reply #32 –
The folks at www.afterdawn.com have picked up on this thread and are telling the world to hold off on 3.96 stable.

I have a feeling this is a problem with an isolated sample....

Lame 3.96 released

Reply #33 –
Afterdawn makes me want to beat myself with a stick.

A sniff of a story about something they know little about causes them to make a page long "qualified" argument.

Yes, even audio compression has its tabloid bitches.

Ruairi
rc55.com - nothing going on

Lame 3.96 released

Reply #34 –
@rc55

I think beating yourself with a stick is going too far.  I think Afterdawn went too far.  There are some talented people around here, but this is a forum, not a scientific journal.

Lame 3.96 released

Reply #36 –
What is wrong with the Afterdawn.com news item?

It says that lame 3.96 is potentially buggy and suggests to wait a few days before applying the latest version to use. The also suggests that the problem may be an isolated case.

This may not be a scientific source, but if someone known another site with LAME related user testing -- please let the world know. Also, last time I check there was a whole lot of whining as the CDFreaks announced the 3.96 to the public - people were in a hurry to alert freaks about the potential bug.

Lame 3.96 released

Reply #37 –
Quote
What is wrong with the Afterdawn.com news item?


First I do not like the title ("LAME project releases a potentially buggy v3.96"). If there are problems in 3.96, it is my sole responsability, and not the one of all LAME developpers. I would have preffered "LAME 3.96 release is potentially buggy"

Quote
3.96 stable introduced a serious bug

You can not call this a bug but a problem. It is not a crash or a programmation error, but an issue regrading psychoacoustic model tuning. I would have preffered "LAME 3.96 release has potential problems/issues"

Quote
According various members of the Hydrogenaudio.org community, the 3.96 causes audible drop outs to the signal.

It is not to THE signal (in general) but to SOME signals (specific samples)

Lame 3.96 released

Reply #38 –
Gabriel,

In the future I will gladly receive your or LAME project's official press announcements, so that you can have your impact in the choice of vocalbulary. If such are not available, I will have to operate based on the fractions of information scattered around the ever unreliable Internet.

Lame 3.96 released

Reply #39 –
Doom9 is also reporting that the new release may be buggy.

Lame 3.96 released

Reply #40 –
I personally think that afterdawn jumped the gun. Just because there is potential for one case of sample degredation - its definately not proof that it is completely buggy. If this mentality was taken then you could safely say that all versions of Lame are buggy because problem samples exist.

They should have just said "3.96 released - lets wait for the reaction to see whether its recommended above 3.90.1". Saying it is potentially buggy when that's not even proven makes a headache for the developers and users. Wait for solid, substantial proof.

eagleray: Not a scientific journal, granted - but this place is certainly a solidly grounded place with people who often know /more/ than what would be mentioned in journals.

Ruairi
rc55.com - nothing going on

Lame 3.96 released

Reply #41 –
Well...I did not mean to get everyone excited.  Especially those of you at CD-RW and its related sites, which are very fine, and I visit them daily.

Does anyone know what the state of this psychoacoustic issue is?  Has it been reproduced on other samples, or systems, or with other users?

Lame 3.96 released

Reply #42 –
Discussion about using -m s with --preset extreme split here.
Let's suppose that rain washes out a picnic. Who is feeling negative? The rain? Or YOU? What's causing the negative feeling? The rain or your reaction? - Anthony De Mello