Looking to do some 320 CBR rips
using the Lame 3.97 Final & Razor lame
want to know if all these are same Settings:
--alt-preset insane
-b 320 -m j -h
-b 320
is there any modified version of the lame 3.97 ? which will give me better quality of files ?
Pls help out
Full Version: Not happy even with 320 CBR :(
Please first learn what ABX is!
The settings are the same.
The settings are the same.
have Huge Files on CBR as well 
read around,VBR seems to be the way to go
wanted to wats the best preset for max Quality in VBR ?
is it
--preset fast extreme
or --preset extreme
read around,VBR seems to be the way to go
wanted to wats the best preset for max Quality in VBR ?
is it
--preset fast extreme
or --preset extreme
By default -q value is 3. So "-b 320" and "-b 320 -m j -h" are not the same. But I doubt that anybody can hear difference between "-b 320 -q 3" and "-b 320 -q 2". Or someone can?
And... vinyas, WHY are you not happy with 320 kbps and want better quality?
And... vinyas, WHY are you not happy with 320 kbps and want better quality?
well i find too lossly to my liking
i know about Flac being lossless,i am trying to encode for a device which supports only mp3 & no flac support
i need to rip about 25 cds
so looking for the best Quality VBR preset .... ?
i know about Flac being lossless,i am trying to encode for a device which supports only mp3
& no flac supporti need to rip about 25 cds
so looking for the best Quality VBR preset .... ?
Perhaps your device is to blame for crappy sound quality and not MP3 itself. Problems that might be causing you dislike the sound are
1) Impedance mismatch and damping factor (impedance bridging) issues between headphone amp and headphones, causing filtering and overloading
2) Clipping caused by DSP, such as volume boost and EQ
3) Weird distortion caused by MP3-specific EQ. Some devices and even some PC software will equalize MP3 by modifying scalefactors, etc. during the decode process. This is usually much less resource intensive (important for small embedded chips and for software originally written for old PC's, e.g. 486)
4) Buggy, non-compliant, or limited accuracy MP3 decoder
1) Impedance mismatch and damping factor (impedance bridging) issues between headphone amp and headphones, causing filtering and overloading
2) Clipping caused by DSP, such as volume boost and EQ
3) Weird distortion caused by MP3-specific EQ. Some devices and even some PC software will equalize MP3 by modifying scalefactors, etc. during the decode process. This is usually much less resource intensive (important for small embedded chips and for software originally written for old PC's, e.g. 486)
4) Buggy, non-compliant, or limited accuracy MP3 decoder
QUOTE(vinyas @ Jun 24 2008, 19:42) 
so looking for the best Quality VBR preset .... ?
For LAME 3.97 it is "-V 0 --vbr-new" or equivalently "--preset fast extreme"
Look at http://wiki.hydrogenaudio.org/index.php?ti...and_source_code
You may even find that LAME 3.97 at "-V 0 --vbr-new" is massive overkill for you. "-V 3 --vbr-new" may be more than sufficient for you and would be a lot smaller. There's no point in wasting bits on something you can't personally hear. 
Cheers, Slipstreem.
Cheers, Slipstreem.
Pardon me if Im wrong, but isnt it so that VBR wouldnt make the quality better (as 320Kb/s is the highest bitrate of the MP3 format anyway) but rather just make the files smaller ?
So to answer the OP's question, AFAIK using VBR wont help you if you feel 320Kb is too limiting. However, and Im sure youll notice, you will have problems convincing people here that you can tell a 320Kb MP3 file from the original WAV.
So to answer the OP's question, AFAIK using VBR wont help you if you feel 320Kb is too limiting. However, and Im sure youll notice, you will have problems convincing people here that you can tell a 320Kb MP3 file from the original WAV.
been testing for an hour now
Settled down for: -V 0 --vbr-new -b 192
seems good
any more tweaks ?
which would be better:
-V 0 --vbr-new -b 192
-V 2 --vbr-new -b 192
wat the difference like ?
Settled down for: -V 0 --vbr-new -b 192
seems good
any more tweaks ?
which would be better:
-V 0 --vbr-new -b 192
-V 2 --vbr-new -b 192
wat the difference like ?
No need to tweak. Just use the recommended settings.
I'm always impressed by people "not happy with 320 CBR" (all those Steve Austin 
).
Home use: use CDs
Away use: you won't hear the difference between 224 (or even 192) and 320...
).Home use: use CDs
Away use: you won't hear the difference between 224 (or even 192) and 320...
QUOTE(vinyas @ Jun 24 2008, 11:33)
been testing for an hour now
Settled down for: -V 0 --vbr-new -b 192
seems good
any more tweaks ?
which would be better:
-V 0 --vbr-new -b 192
-V 2 --vbr-new -b 192
wat the difference like ?
Get rid of the "-b 192" at the end. There is no real point in using it with VBR, unless you want to really handicap and cripple LAME's effectiveness.
-V 0 is a target average bitrate of about ~245 kbps. Some portions of a MP3 will be encoded higher (all the way up to 320kbps), others as low as 32kbps, as LAME sees fit to maintain proper quality.
-V 2 is a target average bitrate of about ~190 kbps. Once again some portions of the MP3 will be encoded at a higher bitrate, other parts as low as 32kbps.
Try Lame 3.98 Beta 8 ... there has a lot been improved.
I was never happy with 3.97, but the 3.98 betas with -V 0 are transparent to me in most cases.
I was never happy with 3.97, but the 3.98 betas with -V 0 are transparent to me in most cases.
QUOTE(WonderSlug @ Jun 24 2008, 12:56)
Get rid of the "-b 192" at the end. There is no real point in using it with VBR, unless you want to really handicap and cripple LAME's effectiveness.
-V 0 is a target average bitrate of about ~245 kbps. Some portions of a MP3 will be encoded higher (all the way up to 320kbps), others as low as 32kbps, as LAME sees fit to maintain proper quality.
-V 2 is a target average bitrate of about ~190 kbps. Once again some portions of the MP3 will be encoded at a higher bitrate, other parts as low as 32kbps.
in -V 0 the bitrates go as low as 128 is that healthy ?
QUOTE(vinyas @ Jun 24 2008, 12:02)
QUOTE(WonderSlug @ Jun 24 2008, 12:56)
Get rid of the "-b 192" at the end. There is no real point in using it with VBR, unless you want to really handicap and cripple LAME's effectiveness.
-V 0 is a target average bitrate of about ~245 kbps. Some portions of a MP3 will be encoded higher (all the way up to 320kbps), others as low as 32kbps, as LAME sees fit to maintain proper quality.
-V 2 is a target average bitrate of about ~190 kbps. Once again some portions of the MP3 will be encoded at a higher bitrate, other parts as low as 32kbps.
in -V 0 the bitrates go as low as 128 is that healthy ?
I've had -V 0 encodings with some frames as low as 32kbps. Those frames were really brief portions of silence in the music (i.e. when a band stops and pauses really quickly for a short time) or a quarter-second of silence at the start or end. There's no real point in encoding those parts any higher. Silence doesn't sound better at 192kbps than it does at 32kbps.
So yes, if LAME determines that a portion sounds transparent at 128kbps and going with a higher bitrate doesn't signficantly improve perceptual quality, let it.
in -V 0 the bitrates go as low as 128 is that healthy ?
QUOTE(vinyas @ Jun 24 2008, 20:07)
in -V 0 the bitrates go as low as 128 is that healthy ?
Yes, for certain sounds such as mono samples, or both channels have sound that is exactly the same (joint stero).
Maybe you should try to do an ABX first. For most samples, I myself find that -V 5 is adequate, especially for portable use.
QUOTE(vinyas @ Jun 25 2008, 03:07)
in -V 0 the bitrates go as low as 128 is that healthy ?
The LAME devs would've disabled that by now if it wasn't. If you can't trust the presets, abx them vs your tweaks and post the results.
As others have said, you should really look into conducting your own blind ABX tests. 320kbps CBR and -V 0 are often considered to be overkill in that their bitrate settings are too high for most people with most samples. Don't let your eyes be the judge (ie looking at the bitrates), let your ears tell you which VBR mode to encode to. There is absolutely no need to encode at -V 0 if your ears can hear up to -V 4 or -V 3. That doesn't mean that you have bad hearing, it just means that you can encode at a lower bitrate thus allowing you to store more songs on your portable player.
There will be a time when you will want to store more than 25 CDs on that thing and using a high bitrate setting (such as 320kbps or -V 0) will really limit your storage. You need to conduct a blind ABX test before your claims can be taken seriously. Other wise, you fall ill to the placebo affect in that you know the bitrate is X and you think that X is low quality.
There will be a time when you will want to store more than 25 CDs on that thing and using a high bitrate setting (such as 320kbps or -V 0) will really limit your storage. You need to conduct a blind ABX test before your claims can be taken seriously. Other wise, you fall ill to the placebo affect in that you know the bitrate is X and you think that X is low quality.
QUOTE(vinyas @ Jun 24 2008, 07:42)
well i find too lossly to my liking
i know about Flac being lossless,i am trying to encode for a device which supports only mp3
& no flac supporti need to rip about 25 cds
so looking for the best Quality VBR preset .... ?
I just wonder what the device is that can play only MP3s and which would provide such output quality to discern a 320 kbit MP3...
The quality of my generic iPod (which only plays MP3's) is such that I can barely if at all hear the difference between 128 (or maybe even 112) and V0. 
With that, and since it's only 1GB, I just transcode my V0's that I want on there down to 128 to save room, then erase them later when I want a new batch of music on it.
With that, and since it's only 1GB, I just transcode my V0's that I want on there down to 128 to save room, then erase them later when I want a new batch of music on it.
For me the best encoder at 320 kbps, CBR, is lame 3.92 (from 3.90 up to 3.93 is fine, I think ). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.
). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.
QUOTE(Juan C. @ Jun 25 2008, 11:28)
For me the best encoder at 320 kbps, CBR, is lame 3.92 (from 3.90 up to 3.93 is fine, I think ). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.
). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.What are those problems? When not referenced, it may mislead users to misinformation!
QUOTE(Egor @ Jun 24 2008, 21:17)
QUOTE(Juan C. @ Jun 25 2008, 11:28)
For me the best encoder at 320 kbps, CBR, is lame 3.92 (from 3.90 up to 3.93 is fine, I think ). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.
). I had some problems with lame encoders from 3.94 encoding at 320 CBR since the lame encoder is now optimized at encoding in VBR mode.What are those problems? When not referenced, it may mislead users to misinformation!
http://www.hydrogenaudio.org/forums/index....st&p=537861
The issue was related to 128 CBR but someone confirmed the issue even at 320 CBR. The sad thing is that previous versions of lame produced better output than the newer ones (3.97+).
QUOTE(Egor @ Jun 25 2008, 00:17)
What are those problems? When not referenced, it may mislead users to misinformation!
Well, I've done an ABX test between 2 files. One was encoded with lame 3.92 and the other was encoded with 3.97. The command line was very simple: -b 320 -ms -h as seen here:
CODE
E:\>lame397 -b320 -ms -h test.wav test397.mp3
LAME 3.97 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 20094 Hz - 20627 Hz
Encoding test.wav to test397.mp3
Encoding as 44.1 kHz 320 kbps stereo MPEG-1 Layer III (4.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
12694/12694 (100%)| 0:43/ 0:43| 0:43/ 0:43| 7.6670x| 0:00
-------------------------------------------------------------------------------
kbps LR % long switch short %
320.0 100.0 95.8 2.5 1.7
Writing LAME Tag...done
ReplayGain: -11.5dB
E:\>lame392 -b320 -ms -h test.wav test392.mp3
LAME version 3.92 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), SIMD, SIMD2
Using polyphase lowpass filter, transition band: 21517 Hz - 22050 Hz
Encoding test.wav to test392.mp3
Encoding as 44.1 kHz 320 kbps stereo MPEG-1 Layer III (4.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
12691/12694 (100%)| 0:21/ 0:21| 0:21/ 0:21| 15.177x| 0:00
average: 320.0 kbps LR: 12694 (100.0%)
Writing LAME Tag...done
E:\>
LAME 3.97 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 20094 Hz - 20627 Hz
Encoding test.wav to test397.mp3
Encoding as 44.1 kHz 320 kbps stereo MPEG-1 Layer III (4.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
12694/12694 (100%)| 0:43/ 0:43| 0:43/ 0:43| 7.6670x| 0:00
-------------------------------------------------------------------------------
kbps LR % long switch short %
320.0 100.0 95.8 2.5 1.7
Writing LAME Tag...done
ReplayGain: -11.5dB
E:\>lame392 -b320 -ms -h test.wav test392.mp3
LAME version 3.92 MMX (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), SIMD, SIMD2
Using polyphase lowpass filter, transition band: 21517 Hz - 22050 Hz
Encoding test.wav to test392.mp3
Encoding as 44.1 kHz 320 kbps stereo MPEG-1 Layer III (4.4x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
12691/12694 (100%)| 0:21/ 0:21| 0:21/ 0:21| 15.177x| 0:00
average: 320.0 kbps LR: 12694 (100.0%)
Writing LAME Tag...done
E:\>
Next I use foobar to do the ABX and here are the results:
CODE
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2008/06/25 17:04:46
File A: E:\test392.mp3
File B: E:\test397.mp3
17:04:46 : Test started.
17:05:38 : 01/01 50.0%
17:06:51 : 02/02 25.0%
17:07:38 : 03/03 12.5%
17:08:16 : 04/04 6.3%
17:08:48 : 04/05 18.8%
17:10:16 : 05/06 10.9%
17:10:30 : 06/07 6.3%
17:10:50 : 07/08 3.5%
17:11:59 : Test finished.
----------
Total: 7/8 (3.5%)
foobar2000 v0.9.4.4
2008/06/25 17:04:46
File A: E:\test392.mp3
File B: E:\test397.mp3
17:04:46 : Test started.
17:05:38 : 01/01 50.0%
17:06:51 : 02/02 25.0%
17:07:38 : 03/03 12.5%
17:08:16 : 04/04 6.3%
17:08:48 : 04/05 18.8%
17:10:16 : 05/06 10.9%
17:10:30 : 06/07 6.3%
17:10:50 : 07/08 3.5%
17:11:59 : Test finished.
----------
Total: 7/8 (3.5%)
As you can see I only failed once. One of the big differences between both files (in which I based my decision) was that 3.97 had less frequency than 3.92 and 3.97 also shows a bit more artifacts. For detecting the artifacts, I used 3D surround (on my speaker).
QUOTE(Juan C. @ Jun 26 2008, 00:29)
As you can see I only failed once. One of the big differences between both files (in which I based my decision) was that 3.97 had less frequency than 3.92 and 3.97 also shows a bit more artifacts. For detecting the artifacts, I used 3D surround (on my speaker).
Mp3 files are not supposed to be used with artificial surround enhancers. We might be interested in your results without such post-processing.
QUOTE(Juan C. @ Jun 26 2008, 00:29)
...As you can see I only failed once. One of the big differences between both files (in which I based my decision) was that 3.97 had less frequency than 3.92 and 3.97 also shows a bit more artifacts. For detecting the artifacts, I used 3D surround (on my speaker).
Whether HF is cut off at 20.5 kHz or at 22 kHz isn't a quality issue. Even young people are hardly expected to be able to hear beyond 20 kHz. It's rather the other way around: mp3 quality can take profit from restricting HF a little bit.
Do you hear the artifacts only when using 3D surround? Did you do your ABXing this way? In this case this is useful only for people who listen the 3D surround way.
Other than that I don't like 3.97 much either because of several issues and do prefer 3.98.
Even when looking at 3.98 it's an open question to me whether it's better than an older version like 3.92, especially at CBR 320. BTW when using 3.92 the -b 320 way instead of alt-preset insane: does that use gpsycho and plain stereo as is done when using 3.90.3?
QUOTE(Juan C. @ Jun 25 2008, 23:29)
For detecting the artifacts, I used 3D surround (on my speaker).
3D surround is probably inverting one of the channels. This is not a normal way of listening. That's why you hear the difference. Lame (and any lossy codec) is optimized for a normal way of listening. So no inverting of channels, no extreme EQ or other weird effects.
QUOTE(halb27 @ Jun 26 2008, 02:41)
Do you hear the artifacts only when using 3D surround? Did you do your ABXing this way? In this case this is useful only for people who listen the 3D surround way.
Maybe the orginal poster is also listening the music this way with some kind of effect and that's why he isn't happy with 320 kbps. It was just an advice.
QUOTE(halb27 @ Jun 26 2008, 02:41)
Whether HF is cut off at 20.5 kHz or at 22 kHz isn't a quality issue.
It depends on the track that you're encoding. In this case I used a song with very high frequency, and the loss of frequency between the 3.92 and 3.97 was noticeable. At least for me. Of course I have done tests with other tracks and I cannot ABX them.
QUOTE(halb27 @ Jun 26 2008, 02:41)
does that use gpsycho and plain stereo as is done when using 3.90.3?
Yes.
QUOTE(Gabriel @ Jun 26 2008, 02:17)
We might be interested in your results without such post-processing.
I will repeat the ABX without any effect enabled.
Done, I've done again the ABX test, this time with the original wav and the lame 3.97 encoded file. All effects were disabled.
The only thing that I know was that file A was the wav and file B was the mp3. Listening music without any enhancer proves that it's really hard to determine which is file A or B. (most of the chances I was guessing).
CODE
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2008/06/26 13:19:24
File A: E:\test.wav
File B: E:\test397.mp3
13:19:24 : Test started.
13:20:55 : 00/01 100.0%
13:22:09 : 00/02 100.0%
13:22:48 : 01/03 87.5%
13:24:44 : 02/04 68.8%
13:25:23 : 03/05 50.0%
13:26:59 : 04/06 34.4%
13:27:54 : 04/07 50.0%
13:28:39 : 05/08 36.3%
13:28:44 : Test finished.
----------
Total: 5/8 (36.3%)
foobar2000 v0.9.4.4
2008/06/26 13:19:24
File A: E:\test.wav
File B: E:\test397.mp3
13:19:24 : Test started.
13:20:55 : 00/01 100.0%
13:22:09 : 00/02 100.0%
13:22:48 : 01/03 87.5%
13:24:44 : 02/04 68.8%
13:25:23 : 03/05 50.0%
13:26:59 : 04/06 34.4%
13:27:54 : 04/07 50.0%
13:28:39 : 05/08 36.3%
13:28:44 : Test finished.
----------
Total: 5/8 (36.3%)
The only thing that I know was that file A was the wav and file B was the mp3. Listening music without any enhancer proves that it's really hard to determine which is file A or B. (most of the chances I was guessing).
QUOTE(Juan C. @ Jun 27 2008, 01:35)
Done, I've done again the ABX test, this time with the original wav and the lame 3.97 encoded file.
I thought you were trying to prove the difference between 3.92 and 3.97, weren't you?
QUOTE(Juan C. @ Jun 27 2008, 01:35)
CODE
Total: 5/8 (36.3%)
The only thing that I know was that file A was the wav and file B was the mp3. Listening music without any enhancer proves that it's really hard to determine which is file A or B. (most of the chances I was guessing).
Maybe I'm wrong but I don't think 5/8 (36.3%) is significant enough.
QUOTE(varoot @ Jun 26 2008, 15:12)
QUOTE(Juan C. @ Jun 27 2008, 01:35)
CODE
Total: 5/8 (36.3%)
The only thing that I know was that file A was the wav and file B was the mp3. Listening music without any enhancer proves that it's really hard to determine which is file A or B. (most of the chances I was guessing).
Maybe I'm wrong but I don't think 5/8 (36.3%) is significant enough.
Indeed, he was saying that he didn't think he could hear the difference, that he was just guessing.
QUOTE(varoot @ Jun 26 2008, 15:12)
QUOTE(Juan C. @ Jun 27 2008, 01:35)
Done, I've done again the ABX test, this time with the original wav and the lame 3.97 encoded file.
I thought you were trying to prove the difference between 3.92 and 3.97, weren't you?
If he can't tell 3.97 from the original then it goes without saying that he can't judge 3.97 to be less transparent than 3.92.
QUOTE(pdq @ Jun 26 2008, 15:19)
If he can't tell 3.97 from the original then it goes without saying that he can't judge 3.97 to be less transparent than 3.92.
I will try again this time with 3.92 and 3.97 with no effects.
CODE
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2008/06/26 15:24:13
File A: E:\test397.mp3
File B: E:\test392.mp3
15:24:13 : Test started.
15:24:59 : 01/01 50.0%
15:26:35 : 02/02 25.0%
15:27:54 : 03/03 12.5%
15:30:29 : 04/04 6.3%
15:31:18 : 04/05 18.8%
15:32:01 : 05/06 10.9%
15:33:15 : 05/07 22.7%
15:34:11 : 05/08 36.3%
15:35:02 : Test finished.
----------
Total: 5/8 (36.3%)
foobar2000 v0.9.4.4
2008/06/26 15:24:13
File A: E:\test397.mp3
File B: E:\test392.mp3
15:24:13 : Test started.
15:24:59 : 01/01 50.0%
15:26:35 : 02/02 25.0%
15:27:54 : 03/03 12.5%
15:30:29 : 04/04 6.3%
15:31:18 : 04/05 18.8%
15:32:01 : 05/06 10.9%
15:33:15 : 05/07 22.7%
15:34:11 : 05/08 36.3%
15:35:02 : Test finished.
----------
Total: 5/8 (36.3%)
Again same results. Please note that I choose a track that is difficut to encode that why I can detect more artifacts on 3.97 than 3.92. In fact, using -ms for an almost mono and clipping track could be challenging to the encoder but I just do this for testing. Maybe someone can run another test to see his/her results.
QUOTE(Juan C. @ Jun 26 2008, 16:45)
CODE
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2008/06/26 15:24:13
15:34:11 : 05/08 36.3%
15:35:02 : Test finished.
----------
Total: 5/8 (36.3%)
foobar2000 v0.9.4.4
2008/06/26 15:24:13
15:34:11 : 05/08 36.3%
15:35:02 : Test finished.
----------
Total: 5/8 (36.3%)
Again same results. Please note that I choose a track that is difficut to encode that why I can detect more artifacts on 3.97 than 3.92.
I must be misunderstanding something, for these ABX results clearly show you can NOT detect more artifacts on 3.97 than on 3.92.
QUOTE(Soap @ Jun 27 2008, 12:01)
I must be misunderstanding something, for these ABX results clearly show you can NOT detect more artifacts on 3.97 than on 3.92.
Yeah, he states he still cannot determine the difference, hence the same results:
QUOTE(Juan C. @ Jun 27 2008, 01:35)
[...] Listening music without any enhancer proves that it's really hard to determine which is file A or B. (most of the chances I was guessing).
QUOTE(Egor @ Jun 27 2008, 00:16)
Yeah, he states he still cannot determine the difference, hence the same results:
Sorry if there has been misinformation here (my english is not the best
) but I yes I cannot tell the difference. Only using the 3D surround I was able to determine which file was each.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.