It was generally the (unofficial) conclusion that fast standard is now better than standard as of 3.97, correct? I'm double-checking, because I need to rip a whole bunch of CDs and I'm impatient for the final 3.97 release.
At the very least, no critical failures were found with fast standard, correct?
Full Version: 3.97: standard or fast standard?
QUOTE(Jebus @ May 24 2005, 04:40 PM)
It was generally the (unofficial) conclusion that fast standard is now better than standard as of 3.97, correct? I'm double-checking, because I need to rip a whole bunch of CDs and I'm impatient for the final 3.97 release.
At the very least, no critical failures were found with fast standard, correct?
At the very least, no critical failures were found with fast standard, correct?
IIRC, "fast" performed slightly better on average in guruboolez tests, but they were very close, so it probably doesn't matter which one you use - there will probably be samples where either one or the other is better. Personally, I'd go for the fast mode, although I am still a bit wary because it is officially in "alpha" status.
Yeah, but it seems like it is more or less consistently better than 3.96.1, and frankly it sounds like the LAME devs aren't working much at all on 3.97 right now (Don't bite me! I'm not saying they have to be working 24/7, just that Gabriel indicated most of the current work is by Takahiro on 4.0).
BTW in 3.97a10: Theres no indicator for --vbr-new? I use: -V 2 --vbr-new and it says:
CODE
'Coriolan & Koenig Stephan' - Ludwig van Beethoven- Symphonie No.6 'Pastorale',
Overtueren 'Coriolan & Koenig Stephan'\0tmp7!1!5.wav
to M:\Desktop\Ludwig van Beethoven- Symphonie No.6 'Pastorale', Overtueren
'Coriolan & Koenig Stephan' - Ludwig van Beethoven- Symphonie No.6 'Pastorale',
Overtueren 'Coriolan & Koenig Stephan'\0tmp7!1!5.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
3850/30944 (12%)| 0:49/ 6:41| 0:49/ 6:41| 2.0130x| 5:51
32 [ 4] *
40 [ 0]
48 [ 0]
56 [ 1] *
64 [ 0]
80 [ 0]
96 [ 0]
112 [ 1] *
128 [ 2] *
160 [ 696] %%%%%%%%%%%%%***
192 [3012] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********
224 [ 134] %%%*
256 [ 0]
320 [ 0]
----------11:47----------------------------------------------------------------
kbps LR MS % long %
187.1 85.8 14.2 100.0
Overtueren 'Coriolan & Koenig Stephan'\0tmp7!1!5.wav
to M:\Desktop\Ludwig van Beethoven- Symphonie No.6 'Pastorale', Overtueren
'Coriolan & Koenig Stephan' - Ludwig van Beethoven- Symphonie No.6 'Pastorale',
Overtueren 'Coriolan & Koenig Stephan'\0tmp7!1!5.mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
3850/30944 (12%)| 0:49/ 6:41| 0:49/ 6:41| 2.0130x| 5:51
32 [ 4] *
40 [ 0]
48 [ 0]
56 [ 1] *
64 [ 0]
80 [ 0]
96 [ 0]
112 [ 1] *
128 [ 2] *
160 [ 696] %%%%%%%%%%%%%***
192 [3012] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%********
224 [ 134] %%%*
256 [ 0]
320 [ 0]
----------11:47----------------------------------------------------------------
kbps LR MS % long %
187.1 85.8 14.2 100.0
AFAIR --vbr-new is now operating without bitrate limits while the old mode still uses a min bitrate of 128kbps.
I've been using the fast mode mostly with preset medium and never noticed any obvious quality compromise.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
QUOTE(dev0 @ May 25 2005, 10:15 AM)
AFAIR --vbr-new is now operating without bitrate limits while the old mode still uses a min bitrate of 128kbps.
I've encoded some files using LAME 3.97a10 & --preset standard and there wasn't any bitrate limit either...
It's cool to see so much continued enthusiasm for LAME development!
Hopefully, the stable version will release soon
Hopefully, the stable version will release soon
QUOTE(beto @ May 25 2005, 01:18 PM)
I've been using the fast mode mostly with preset medium and never noticed any obvious quality compromise.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
This post appears to be in violation of rule #8.
QUOTE(Synaptic Line Noise @ May 25 2005, 06:23 PM)
QUOTE(beto @ May 25 2005, 01:18 PM)
I've been using the fast mode mostly with preset medium and never noticed any obvious quality compromise.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
This post appears to be in violation of rule #8.
Oh settle down, he didn't claim to hear a difference, he claimed to NOT hear one. You don't have to back up non-claims with evidence, otherwise no one would ever get anything done!
QUOTE(Synaptic Line Noise @ May 25 2005, 09:23 PM)
QUOTE(beto @ May 25 2005, 01:18 PM)
I've been using the fast mode mostly with preset medium and never noticed any obvious quality compromise.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
This post appears to be in violation of rule #8.
IMHO as there are no claims regarding audio quality but rather a statement that it is good enough for him, there is no violation of TOS 8.
To clarify some things.
I've noticed during tests that lame --vbr-new alpha produces on average better sound than usual VBR mode. Alpha 5, 6 & 7 with -V5, -V4 & -V2 settings were affected. 3.96.1 also share the same issues than these alpha (see this additionnal listening test).
But Gabriel has updated the defaulted VBR mode with alpha 8 whereas --vbr-new was unchanged. I've evaluated these fresh tunings with -V3 & -V2 settings, and it reveals nice progress. Therefore, the gap between --vbr-new and default VBR was reduced, or could even be completely filled since alpha 8. Is someone interested to ABX directly both VBR modes with latests alphas?
DIRECT LINKS:
• alpha 5 -V4 - 20 samples and alpha 5 -V5 - 6 samples
• alpha 6 -V2 - 20 samples
• alpha 7 -V4 - 54 samples
• alpha 8 -V3 - 20 samples and alpha 8 -V2 - 20 samples
EDIT: --vbr-old & --vbr-new started to be updated with alpha 7, up to -V4. Tunings were followed with -V3 & -V2, but only since alpha 8 (as mentioned first). Improvements were serious, and my -V4 tests on 54 samples revealed that --vbr-new has still some advantages over defaulted mode (especially with some critical samples: microattacks issues).
I've noticed during tests that lame --vbr-new alpha produces on average better sound than usual VBR mode. Alpha 5, 6 & 7 with -V5, -V4 & -V2 settings were affected. 3.96.1 also share the same issues than these alpha (see this additionnal listening test).
But Gabriel has updated the defaulted VBR mode with alpha 8 whereas --vbr-new was unchanged. I've evaluated these fresh tunings with -V3 & -V2 settings, and it reveals nice progress. Therefore, the gap between --vbr-new and default VBR was reduced, or could even be completely filled since alpha 8. Is someone interested to ABX directly both VBR modes with latests alphas?
DIRECT LINKS:
• alpha 5 -V4 - 20 samples and alpha 5 -V5 - 6 samples
• alpha 6 -V2 - 20 samples
• alpha 7 -V4 - 54 samples
• alpha 8 -V3 - 20 samples and alpha 8 -V2 - 20 samples
EDIT: --vbr-old & --vbr-new started to be updated with alpha 7, up to -V4. Tunings were followed with -V3 & -V2, but only since alpha 8 (as mentioned first). Improvements were serious, and my -V4 tests on 54 samples revealed that --vbr-new has still some advantages over defaulted mode (especially with some critical samples: microattacks issues).
Would you conclude from this that from 3.97, --preset medium should be mapped to -V4 --vbr-new. I assume that --preset medium should yield the best quality possible using -V4 (and it's target bitrate), independent from speed achieved.
QUOTE(Jebus @ May 26 2005, 11:39 AM)
QUOTE(Synaptic Line Noise @ May 25 2005, 06:23 PM)
QUOTE(beto @ May 25 2005, 01:18 PM)
I've been using the fast mode mostly with preset medium and never noticed any obvious quality compromise.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
I never tried to ABX against --vbr-old because quality is good enough for me and the speed increase really does pay off.
This post appears to be in violation of rule #8.
Oh settle down, he didn't claim to hear a difference, he claimed to NOT hear one. You don't have to back up non-claims with evidence, otherwise no one would ever get anything done!
One more thing. You shouldn't ABX encodes against each other; you should ABX them with the original. You can use ABC/HR for comparing encodes.
QUOTE(tycho @ May 29 2005, 12:27 PM)
Would you conclude from this that from 3.97, --preset medium should be mapped to -V4 --vbr-new.
From my own experience, yes. But it's only mine, and it would be really nice to see additional feedback, coming from different people, tastes, hardware, etc...
QUOTE(kjoonlee @ May 29 2005, 01:51 PM)
One more thing. You shouldn't ABX encodes against each other; you should ABX them with the original. You can use ABC/HR for comparing encodes.
Why not?
Let's take an example: Encode to 50kbps two times with a tiny change of some setting in the encoder. Now to decide how this changed parameter influences the encoding, it's not enough to just conclude that you can hear a difference between the original and to each of the 50kbps encoded files. You also need to determine that you can actually hear a difference between the two encoded files, and this is done by - tada - ABX. Of course abc/hr can help you do this, but you could use any other tool as well. Now, can you explain to me why one shouldn't ABX encodes against eachother?
Oh well, I guess ABX is useful in that case, but that doesn't tell you which one is closer to the original.
You shouldn't ABX encodes against each other if you want to draw conclusions about encoding quality, IMHO.
You shouldn't ABX encodes against each other if you want to draw conclusions about encoding quality, IMHO.
QUOTE(kjoonlee @ May 29 2005, 01:13 PM)
You shouldn't ABX encodes against each other if you want to draw conclusions about encoding quality, IMHO.
From my own experience, comparing directly two encodings is a very important step if I really want to know if both encodings really sounds different.
ABC/HR step is generally not enough. It's easy to give two different notes to two different encodings, but if you must know if the difference really exists, the direct ABXing is necessary.
ABXing reference vs encoding is generally easier than ABXing encoding vs encoding. It's especially true when you try to compare encodings that are really close (ex: lame --standard vs lame -fast-standard).
You should also keep in mind that reference is still accessible if needed during the ABX comparison between two different encodings. At least with java ABC/HR.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.