Hi and sorry for my bad English
The DEFINITIVE question is:
TODAY, for the BETTER possible results in "--alt-preset standard" WHAT encoder is better on a large number of audio tracks ? ( I know very well that on one song can be better encoder "A" and on another song encoder "B")
Lame 3.90.3 using "--alt-preset standard
OR
Lame 3.97 alpha 1 using "V2"
OR
Lame 3.97 alpha 1 using "V2 NEW VBR"
Thanks
Full Version: What Encoder since NOW
Lame 3.97 alpha 1 using "V2 NEW VBR" is an obvious answer, but...
If you don't hear the difference between the encoders, there is no point to be doubtful about it.
If you don't hear the difference between the encoders, there is no point to be doubtful about it.
QUOTE (pecosbill76 @ Sep 22 2005, 11:14 AM)
Hi and sorry for my bad English
The DEFINITIVE question is:
TODAY, for the BETTER possible results in "--alt-preset standard" WHAT encoder is better on a large number of audio tracks ? ( I know very well that on one song can be better encoder "A" and on another song encoder "B")
Lame 3.90.3 using "--alt-preset standard
OR
Lame 3.97 alpha 1 using "V2"
OR
Lame 3.97 alpha 1 using "V2 NEW VBR"
Thanks
The DEFINITIVE question is:
TODAY, for the BETTER possible results in "--alt-preset standard" WHAT encoder is better on a large number of audio tracks ? ( I know very well that on one song can be better encoder "A" and on another song encoder "B")
Lame 3.90.3 using "--alt-preset standard
OR
Lame 3.97 alpha 1 using "V2"
OR
Lame 3.97 alpha 1 using "V2 NEW VBR"
Thanks
Why use Lame 3.97 alpha 1 when there were numerous alpha versions (and now a beta version) that have come after it?
Lame 3.97b1 (not alpha 1) seems like it has better quality at -V 2 --vbr-new than -V 2. However, I don't know how it pairs up against Lame 3.90.3 or Lame 3.96.1
From my testings on my car system (don't know the specifics but a Sony receiver, 4 rockford fausgate 4-way 500 watt speakers, and 2 rocksford fausgate 1500 watt subwoofers all being pushed by 5 amps) that I recently put in my car, I could hear that Lame 3.97b1 had an increased sound quality (at -V 2 --vbr-new) when comparing it to Lame 3.96.1 and Lame 3.90.3 at the preset standard setting. Keep in mind that this is just with my equipment, my ears, and I tried to do it as blind as possible. I still don't know the true specs of my system as my friend ordered them for me and helped me install the system. Additionally, there could be flaws in my audio equipment. I am just trying not to violate TOS#8.
From my testings on my car system (don't know the specifics but a Sony receiver, 4 rockford fausgate 4-way 500 watt speakers, and 2 rocksford fausgate 1500 watt subwoofers all being pushed by 5 amps) that I recently put in my car, I could hear that Lame 3.97b1 had an increased sound quality (at -V 2 --vbr-new) when comparing it to Lame 3.96.1 and Lame 3.90.3 at the preset standard setting. Keep in mind that this is just with my equipment, my ears, and I tried to do it as blind as possible. I still don't know the true specs of my system as my friend ordered them for me and helped me install the system. Additionally, there could be flaws in my audio equipment. I am just trying not to violate TOS#8.
QUOTE (kornchild2002 @ Sep 22 2005, 07:22 PM)
Lame 3.97b1 (not alpha 1) seems like it has better quality at -V 2 --vbr-new than -V 2. However, I don't know how it pairs up against Lame 3.90.3 or Lame 3.96.1
From my testings on my car system (don't know the specifics but a Sony receiver, 4 rockford fausgate 4-way 500 watt speakers, and 2 rocksford fausgate 1500 watt subwoofers all being pushed by 5 amps) that I recently put in my car, I could hear that Lame 3.97b1 had an increased sound quality (at -V 2 --vbr-new) when comparing it to Lame 3.96.1 and Lame 3.90.3 at the preset standard setting. Keep in mind that this is just with my equipment, my ears, and I tried to do it as blind as possible. I still don't know the true specs of my system as my friend ordered them for me and helped me install the system. Additionally, there could be flaws in my audio equipment. I am just trying not to violate TOS#8.
From my testings on my car system (don't know the specifics but a Sony receiver, 4 rockford fausgate 4-way 500 watt speakers, and 2 rocksford fausgate 1500 watt subwoofers all being pushed by 5 amps) that I recently put in my car, I could hear that Lame 3.97b1 had an increased sound quality (at -V 2 --vbr-new) when comparing it to Lame 3.96.1 and Lame 3.90.3 at the preset standard setting. Keep in mind that this is just with my equipment, my ears, and I tried to do it as blind as possible. I still don't know the true specs of my system as my friend ordered them for me and helped me install the system. Additionally, there could be flaws in my audio equipment. I am just trying not to violate TOS#8.
Sorry again for my bad English:
Is there a BLIND TEST done on a great number of audio tracks which can sure tell: " 3.97 beta 1 is better than 3.90.3" at V2 (or equivalent) setting ?
QUOTE (pecosbill76 @ Sep 23 2005, 08:01 AM)
Is there a BLIND TEST done on a great number of audio tracks which can sure tell: " 3.97 beta 1 is better than 3.90.3" at V2 (or equivalent) setting ?
There has not been any big testing on that question. Many tests were done on 3.96.1 vs 3.90.3, with the result that 3.96.1 was judged the best for lower bitrates. See here for the whole test. The tests for 3.90.3 aps vs 3.96 ps were less definitive, but most of them favored 3.96. I'm not sure why 3.90.3 is still the recommended version honestly. IMHO when 3.97 is final there should be a test done vs 3.90.3, but the old one should have to defend itself, ie the default outcome at this point should be to update the recommended version.
I think it is almost certain that 3.97 has better quality at all bitrates than 3.90.3, so you should choose the new one. Especially with the reports of better quality using vbr-new,: that means there are two "steps" where quality is likely to increase.
The 3.97 alphas HAVE been tested against 3.90.3, and on a vast majority of samples, performed better than the latter, on various bitrates. Search for listening tests by guruboolez. What you should also know, is that the vbr engine hasn't changed in quite a few versions now, so it's basically the same in the beta as in the last alphas (which had been tested)
QUOTE (Klyith @ Sep 23 2005, 01:35 PM)
QUOTE (pecosbill76 @ Sep 23 2005, 08:01 AM)
Is there a BLIND TEST done on a great number of audio tracks which can sure tell: " 3.97 beta 1 is better than 3.90.3" at V2 (or equivalent) setting ?
...I'm not sure why 3.90.3 is still the recommended version honestly. ... ie the default outcome at this point should be to update the recommended version.
...
err, if you look at List of recomemnded settings topic, you find, that lame 3.97beta1 is already the recommended version.
The other 2 sticky topics are of older age, therefore you have always a date given of latest update of the stickies.
Those 2 stickies about compile versions need to be updated now, maybe a moderator could do so.
I tried to answer this question for myself last week by using ABC/HR Java between
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
Hi from Italy.
I tried 3.90.3 aps, 3.97b1 -V 2 and -V 2 --vbr-new on a random choose of 10 audio tracks: rock and pop, studio and live recordings, from 70's to 2005.
ON ALL SETTING I ADDES THE --scale 0.95 OPTION TO AVOID VIRTUALLY ANY CLIPPING.
First impressions using my EYES and NOT my ears:
-V 2 --vbr-new is 100% faster than 3.90.3 aps during encoding, but 3.97b1 -V 2 is faster too of about 50% compared to 3.90.3 aps.
The filesize of -V 2 --vbr-new is the smaller one: 3.90.3 aps is 5.1% bigger and 3.97b1 -V 2 is 2% bigger
The minimum bitrate on 3.90.3 aps is 128 kbps (or 32kbps on analog silence).
Encoder 3.97 b1 in very few blocks uses 112 kbps, 96 kbps and even smaller bitrates.
I tried 3.90.3 aps, 3.97b1 -V 2 and -V 2 --vbr-new on a random choose of 10 audio tracks: rock and pop, studio and live recordings, from 70's to 2005.
ON ALL SETTING I ADDES THE --scale 0.95 OPTION TO AVOID VIRTUALLY ANY CLIPPING.
First impressions using my EYES and NOT my ears:
-V 2 --vbr-new is 100% faster than 3.90.3 aps during encoding, but 3.97b1 -V 2 is faster too of about 50% compared to 3.90.3 aps.
The filesize of -V 2 --vbr-new is the smaller one: 3.90.3 aps is 5.1% bigger and 3.97b1 -V 2 is 2% bigger
The minimum bitrate on 3.90.3 aps is 128 kbps (or 32kbps on analog silence).
Encoder 3.97 b1 in very few blocks uses 112 kbps, 96 kbps and even smaller bitrates.
QUOTE (singaiya @ Sep 23 2005, 01:30 PM)
I tried to answer this question for myself last week by using ABC/HR Java between
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
What was your reference when testing?
QUOTE (Shade[ST] @ Sep 23 2005, 01:50 PM)
QUOTE (singaiya @ Sep 23 2005, 01:30 PM)
I tried to answer this question for myself last week by using ABC/HR Java between
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
3.90.3 --alt-preset standard
3.96.1 -V 2
3.97b1 -V 2 --vbr-new
And I was highly surprised to see in the logs that I ranked 3.90.3 slightly above the others both times. I'll post the details when I get home. Still not sure that I did everything correctly as this was my first time using the ABC software. And IIRC I only tried it on a few "killers" because it's hard for me to ABX much of anything at this bitrate level, but I have successful ABX logs on the files in question. Again, will post those when I'm at home.
What was your reference when testing?
Well, after looking at my logs I see that I indeed made some errors -- I guess it's to be expected with a first time
. Just learning to set up ABC/HR took longer than I expected 
The reference file (for the only valid test I did) was castanets.In short, what I found was that my previous successful ABX sessions on 3.97 samples of fatboy and kraftwerk were at -V 2, not -V 2 --vbr-new. Tonight I tried ABXing 3.97 -V 2 --vbr-new and failed on both those samples (but I did succeed with castanets). I also failed to ABX 3.90.3 --aps with fatboy, and once with castanets (but once I passed 8/8).
It's my understanding that an ABC/HR session is valid when all ABCs have successful ABX's to match. So with that criterion, then my only valid one is with castanets, where twice I ranked 3.90.3 highest. My ABC tests ranking 3.90.3 higher on the other samples are invalid because I could not ABX them against 3.97 -V2 --vbr-new, the new vbr being the reason for the ABC test in the first place.
In conclusion, 3.90.3 sounds slightly better on castanets *to me*. But they were all ranked with good marks, ie, not annoying. And on the other samples, I could not ABX 3.97 -V 2 --vbr-new (though I did with --old).
Here's the logs:
CODE
ABC/HR for Java, Version 0.5b, 23 September 2005
Testname: castanets2
Tester: lame versions castanets2
1L = C:\My Documents\Min\test files\castanets_3.90.3.mp3
2L = C:\My Documents\Min\test files\castanets_3.96.1.mp3
3R = C:\My Documents\Min\test files\castanets_3.97b1.mp3
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\My Documents\Min\test files\castanets_3.90.3.mp3
1L Rating: 4.5
1L Comment:
---------------------------------------
2L File: C:\My Documents\Min\test files\castanets_3.96.1.mp3
2L Rating: 4.3
2L Comment:
---------------------------------------
3R File: C:\My Documents\Min\test files\castanets_3.97b1.mp3
3R Rating: 4.0
3R Comment:
---------------------------------------
ABX Results:
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
8 out of 8, pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.96.1.mp3
8 out of 8, pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.90.3.mp3
8 out of 8, pval = 0.0030
---- Detailed ABX results ----
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
Playback Range: 00.000 to 06.144
8:00:19 PM p 1/1 pval = 0.5
8:00:24 PM p 2/2 pval = 0.25
8:00:28 PM p 3/3 pval = 0.125
8:00:30 PM p 4/4 pval = 0.062
8:00:37 PM p 5/5 pval = 0.031
8:00:41 PM p 6/6 pval = 0.015
8:00:44 PM p 7/7 pval = 0.0070
8:00:48 PM p 8/8 pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.96.1.mp3
Playback Range: 00.000 to 06.144
7:58:49 PM p 1/1 pval = 0.5
7:58:53 PM p 2/2 pval = 0.25
7:58:59 PM p 3/3 pval = 0.125
7:59:09 PM p 4/4 pval = 0.062
7:59:12 PM p 5/5 pval = 0.031
7:59:20 PM p 6/6 pval = 0.015
7:59:29 PM p 7/7 pval = 0.0070
7:59:38 PM p 8/8 pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.90.3.mp3
Playback Range: 00.000 to 06.144
7:57:15 PM p 1/1 pval = 0.5
7:57:23 PM p 2/2 pval = 0.25
7:57:30 PM p 3/3 pval = 0.125
7:57:44 PM p 4/4 pval = 0.062
7:57:47 PM p 5/5 pval = 0.031
7:57:53 PM p 6/6 pval = 0.015
7:57:59 PM p 7/7 pval = 0.0070
7:58:04 PM p 8/8 pval = 0.0030
Testname: castanets2
Tester: lame versions castanets2
1L = C:\My Documents\Min\test files\castanets_3.90.3.mp3
2L = C:\My Documents\Min\test files\castanets_3.96.1.mp3
3R = C:\My Documents\Min\test files\castanets_3.97b1.mp3
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\My Documents\Min\test files\castanets_3.90.3.mp3
1L Rating: 4.5
1L Comment:
---------------------------------------
2L File: C:\My Documents\Min\test files\castanets_3.96.1.mp3
2L Rating: 4.3
2L Comment:
---------------------------------------
3R File: C:\My Documents\Min\test files\castanets_3.97b1.mp3
3R Rating: 4.0
3R Comment:
---------------------------------------
ABX Results:
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
8 out of 8, pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.96.1.mp3
8 out of 8, pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.90.3.mp3
8 out of 8, pval = 0.0030
---- Detailed ABX results ----
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
Playback Range: 00.000 to 06.144
8:00:19 PM p 1/1 pval = 0.5
8:00:24 PM p 2/2 pval = 0.25
8:00:28 PM p 3/3 pval = 0.125
8:00:30 PM p 4/4 pval = 0.062
8:00:37 PM p 5/5 pval = 0.031
8:00:41 PM p 6/6 pval = 0.015
8:00:44 PM p 7/7 pval = 0.0070
8:00:48 PM p 8/8 pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.96.1.mp3
Playback Range: 00.000 to 06.144
7:58:49 PM p 1/1 pval = 0.5
7:58:53 PM p 2/2 pval = 0.25
7:58:59 PM p 3/3 pval = 0.125
7:59:09 PM p 4/4 pval = 0.062
7:59:12 PM p 5/5 pval = 0.031
7:59:20 PM p 6/6 pval = 0.015
7:59:29 PM p 7/7 pval = 0.0070
7:59:38 PM p 8/8 pval = 0.0030
Original vs C:\My Documents\Min\test files\castanets_3.90.3.mp3
Playback Range: 00.000 to 06.144
7:57:15 PM p 1/1 pval = 0.5
7:57:23 PM p 2/2 pval = 0.25
7:57:30 PM p 3/3 pval = 0.125
7:57:44 PM p 4/4 pval = 0.062
7:57:47 PM p 5/5 pval = 0.031
7:57:53 PM p 6/6 pval = 0.015
7:57:59 PM p 7/7 pval = 0.0070
7:58:04 PM p 8/8 pval = 0.0030
earlier ABC session without ABXing on same clip:
CODE
ABC/HR for Java, Version 0.5b, 17 September 2005
Testname: lame versions
Tester: lame versions castanets
1L = C:\My Documents\Min\test files\castanets_3.97b1.mp3
2R = C:\My Documents\Min\test files\castanets_3.90.3.mp3
3L = C:\My Documents\Min\test files\castanets_3.96.1.mp3
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\My Documents\Min\test files\castanets_3.97b1.mp3
1L Rating: 3.0
1L Comment: slightly smeared sounding
---------------------------------------
2R File: C:\My Documents\Min\test files\castanets_3.90.3.mp3
2R Rating: 4.0
2R Comment: harder to tell, barely perceived and not annoying
---------------------------------------
3L File: C:\My Documents\Min\test files\castanets_3.96.1.mp3
3L Rating: 3.5
3L Comment:
---------------------------------------
ABX Results:
Testname: lame versions
Tester: lame versions castanets
1L = C:\My Documents\Min\test files\castanets_3.97b1.mp3
2R = C:\My Documents\Min\test files\castanets_3.90.3.mp3
3L = C:\My Documents\Min\test files\castanets_3.96.1.mp3
---------------------------------------
General Comments:
---------------------------------------
1L File: C:\My Documents\Min\test files\castanets_3.97b1.mp3
1L Rating: 3.0
1L Comment: slightly smeared sounding
---------------------------------------
2R File: C:\My Documents\Min\test files\castanets_3.90.3.mp3
2R Rating: 4.0
2R Comment: harder to tell, barely perceived and not annoying
---------------------------------------
3L File: C:\My Documents\Min\test files\castanets_3.96.1.mp3
3L Rating: 3.5
3L Comment:
---------------------------------------
ABX Results:
QUOTE (singaiya @ Sep 23 2005, 10:30 PM)
Still not sure that I did everything correctly as this was my first time using the ABC software
You should do an additional ABX between 3.90.3 and 3.97 to ensure positive (ehm, not quite positive in this case…) results.
QUOTE (Mo0zOoH @ Sep 24 2005, 08:11 AM)
QUOTE (singaiya @ Sep 23 2005, 10:30 PM)
Still not sure that I did everything correctly as this was my first time using the ABC software
You should do an additional ABX between 3.90.3 and 3.97 to ensure positive (ehm, not quite positive in this case…) results.
The positive ABX for castanets is at the end of the first codebox, positive for all three versions under test.
what I meant, was, did you test these againt a wave file? It doesn't appear so in the log...
Edit : what i mean, you need an original reference file, lossless, to which you can compare, because all three mp3 encoding can have some type of defect. It means nothing that you can differentiate two encodings, say 3.90.3 from 3.97, since one of them may sound nearer to the original, and you don't know which.
Edit : what i mean, you need an original reference file, lossless, to which you can compare, because all three mp3 encoding can have some type of defect. It means nothing that you can differentiate two encodings, say 3.90.3 from 3.97, since one of them may sound nearer to the original, and you don't know which.
I mean 3.97 vs. 3.90.3, not each of them against original.
The difference in ranking is based on the difference in sound (right?), so if you are really able to hear one, you should get a positive ABX result.
The difference in ranking is based on the difference in sound (right?), so if you are really able to hear one, you should get a positive ABX result.
QUOTE (Shade[ST] @ Sep 24 2005, 09:59 AM)
what I meant, was, did you test these againt a wave file? It doesn't appear so in the log...
Edit : what i mean, you need an original reference file, lossless, to which you can compare, because all three mp3 encoding can have some type of defect. It means nothing that you can differentiate two encodings, say 3.90.3 from 3.97, since one of them may sound nearer to the original, and you don't know which.
Edit : what i mean, you need an original reference file, lossless, to which you can compare, because all three mp3 encoding can have some type of defect. It means nothing that you can differentiate two encodings, say 3.90.3 from 3.97, since one of them may sound nearer to the original, and you don't know which.
Of course I tested against all three encoded versions against lossless. I definitely understand the need to do that. But the logs of ABC/HR don't necessarily tell that, they say
CODE
---- Detailed ABX results ----
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
Original vs C:\My Documents\Min\test files\castanets_3.97b1.mp3
Where "original" is the input file I designated, which was "C:\My Documents\Min\test files\castanets.wv". Perhaps the author of ABC/HR could make the logfile write the full path of what the "original" file chosen was, but there's no other way for me to prove that it was a lossless version I was testing against.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.