Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Alex B]

Right now I really only have one device, but we are thinking about getting a small flash based device for activities that would be too much for a HD based device.  Given that it would be non-critical listening and space is a concern I would use CBR or VBR @ ~128kbps.

For my big device (currently 60GB) I use VBR High Fast (in MC terminology).

I would recommend trying LAME -V 5 --vbr-new. It is already usually almost transparent with most music material. The average bitrate is normally about 130-135 kbps with pop/rock etc. Classical and other less intense genres can result a bit lower bitrates. Genres like Speed Metal may use more than 135 kbps.

If the quality is good enough then only one cache archive is needed for all portables.

Things may be different if you use an iPod with a dedicated headphone amp & high end cans in a quiet place for critical listening. Then perhaps a higher VBR quality level would be better.


Actually JRiver could add this setting to the VBR quality options and name it as "Portable". And perhaps also one setting below it for those who need to squeeze files even more.

The new options could be then:

Small Portable -V 6 --vbr-new
HQ Portable -V 5 --vbr-new
Normal -V 4 --vbr-new  (= --preset fast medium, this is directly mapped inside LAME)
High -V 2 --vbr-new  (= --preset fast standard, this is directly mapped inside LAME)
Extreme -V 0 --vbr-new  (= --preset fast extreme, this is directly mapped inside LAME)

The Fast option (--vbr-new) is the currently recommended mode (by the LAME developers and HA). The LAME developers have done a lot of work for making it as good or better than the old VBR mode. According to the recent test results they have succeeded. Currently the Fast mode is generally as good as the old mode and often better. Only a few minor problems have been reported.

The Fast switch could be removed and the --vbr-new mode could be the default mode. At least the now incorrect "much faster, but slightly lower quality" text should be removed.

Those who want to tweak LAME options can continue using the custom text box.

[glynor]

Actually JRiver could add this setting to the VBR quality options and name it as "Portable". And perhaps also one setting below it for those who need to squeeze files even more.

Small Portable -V 6 --vbr-new
HQ Portable -V 5 --vbr-new
Normal -V 4 --vbr-new  (= --preset fast medium, this is directly mapped inside LAME)
High -V 2 --vbr-new  (= --preset fast standard, this is directly mapped inside LAME)
Extreme -V 0 --vbr-new  (= --preset fast extreme, this is directly mapped inside LAME)

That's a FABULOUS idea!  If there were an election, I'd vote for it.

[Jakester]

Hi Alex B - thanks for all your contributions to this board, you always have the latest info on encoding qualities and it's great that you pass the info on to all of us.

Quick question - I'm currently running MC 11.1.194 - what version of LAME does it use?  Is the "fast" option better than the normal speed in this version of MC?

If I upgrade to the latest MC 12 I assume the latest LAME comes with it?

Thanks.

[Alex B]

The fast option is fine to use with MC11.1 and MC12. It is often about 1.3 - 1.4 x faster without any quality problems.


Both MC versions use LAME 3.97 beta 2. LAME 3.97 beta 3 was released in August 2006. It contained only one small fix:

Just a small fix over beta2 regarding a potential problem with some specific signals (like trumpet)

The current version is LAME 3.97 (not a beta anymore). It was officially relesed only a couple of weeks ago. The only difference to beta 3 is in labelling:

This version is identical to 3.97b3, which was promoted to release.

The quotes are from the LAME release notes: http://sourceforge.net/project/showfiles.php?group_id=290&package_id=309


Surely JRiver will update the used LAME version soon. I have not bothered to ask for this separately since I have had some other MP3 encoding related things in my list (mainly this "portable" VBR setting request).

The release version of LAME 3.97 (exe for PCs) is available e.g. here: http://www.rarewares.org/dancer/dancer.php?f=109

The source code link is: http://sourceforge.net/project/showfiles.php?group_id=290

[Jakester]

Great!  Thanks Alex.

[johnnyboy]

Hey Alex,
Just the man I was after by the sounds of it

I ripped a ton of my albums and was leaving them as APE because I wanted as near to perfect as possible for the files.

Considering the fact I'm using some less than $300 equipment to play them back, how close to perfect can mp3 with LAME get in terms of audio difference on that type of equipment, not in terms of there is obviously technically some data loss.
What's the absolutely highest quality settings to use with LAME at the moment?

Thanks

[BartMan01]

Small Portable -V 6 --vbr-new
HQ Portable -V 5 --vbr-new
Normal -V 4 --vbr-new  (= --preset fast medium, this is directly mapped inside LAME)
High -V 2 --vbr-new  (= --preset fast standard, this is directly mapped inside LAME)
Extreme -V 0 --vbr-new  (= --preset fast extreme, this is directly mapped inside LAME)

I like that.  While in there, can we also see the current 'standard' switches at the bottom of that screen.  When you look at the recommended settings they are usually expressed as 'Normal -V 4 --vbr-new' instead of the outdated '--preset fast medium'.

[lalittle]

I've seen some interesting reports regarding the "fast" option.  Is this the same as the "new" option (as some reports indicate)?  Other than speed and size, is there any reason to use the "new" or "fast" option -- i.e. does the normal option off any advantages?  Does the normal option still "potentially" have an advantage in quality?  I'm really not that concerned with speed -- my choice will be based purely on quality.

Thanks,

Larry

[Alex B]

I've seen some interesting reports regarding the "fast" option.

What exactly are these interesting reports?

Is this the same as the "new" option (as some reports indicate)?  Other than speed and size, is there any reason to use the "new" or "fast" option -- i.e. does the normal option off any advantages?  Does the normal option still "potentially" have an advantage in quality?  I'm really not that concerned with speed -- my choice will be based purely on quality.

- I already said: --vbr-new = the old fast presets
- the average size is practically identical.
- according to my tests the new mode is usually 1.3 - 1.4 x faster
- the quality can be debated, but in any case the new mode is not usually worse. It should be more often slightly better.

BTW, the new mode is not exactly "new". This VBR mode has been included in LAME since late 3.8x betas. When 3.90 was released (2001) the LAME developers said that it should be on par with the "old" VBR mode. However, a bit more problems were found with the new mode at that time.

Certain Hydrogen Audio members who obviously have those legendary golden ears agreed that in LAME 3.97 the new option was finally generally better. This was about a year ago during the LAME 3.97 alpha stage.

We mortals can only try to guess which one is better. The difference is not audible to me in most cases.

Have you ever tried to ABX test LAME VBR vs. the original? It is already difficult at -V 5. At -V 0 I can distinguish only certain problem samples from the lossless source files. This is by using ABC/HR blind test software and high-end headphones. I suppose that usually it is impossible to make any difference between the new and old VBR mode in casual listening.

[lalittle]

What exactly are these interesting reports?

Basically, that the "fast" switch can create higher quality results in "some" cases.  This is what I find interesting -- i.e. the idea that it isn't "necessarily" higher quality, and that the previous encoding might actually be better in some cases.

- I already said: --vbr-new = the old fast presets

Thanks -- I missed that.

- the quality can be debated, but in any case the new mode is not usually worse. It should be more often slightly better.

It's the "not usually" that gives me pause.  If the idea was to create a "faster" encoder, it seems like some corners might have to be cut.  The idea that a "fast" switch would create higher quality is counter-intuitive to me.  Is the name simply misleading?

Have you ever tried to ABX test LAME VBR vs. the original?

Once in a while I do this sort of thing, but it's easy to get caught up in it, which gets annoying.  This is why I'm interested to hear what the "golden ear" crowd has to say -- it allows me to be lazy in this regard.

Thanks again,

Larry

[Alex B]

The idea that a "fast" switch would create higher quality is counter-intuitive to me.  Is the name simply misleading?

It is called progress.

The developers have tweaked the new mode for the last several years.

MP3 is an old format in this digital era and the LAME developers have tried to keep it compentive in terms of speed and quality.

They have done surprisingly good job. For example LAME 3.97 b2 -V 5 --vbr-new was on par with the latest modern encoders in a recent public 128 kbps multi-format listening test at HA. The results are here: http://www.maresweb.de/listening-tests/mf-128-1/results.htm

Guruboolez from HA tested LAME 3.97 alpha 5 new vs old mode at -V 2 here:
http://www.hydrogenaudio.org/forums/index.php?s=250f3fcea6d94e7c1cd781237783b1bf&showtopic=30631&view=findpost&p=266785
After that seven alpha and two beta versions were tweaked before the now released v. 3.97.

This quote is from HA's LAME wiki:

The --vbr-new switch enables the new VBR mode. Lame will encode much faster than the old/default VBR mode. In terms of quality, --vbr-new appears to be better than the old model, but reports of artifacts when using the new model do exist. Despite these possible issues, --vbr-new is currently recommended over the default VBR mode due to both the speed and quality increases afforded by the new algorithm.

Edit

Another test made by guruboolez during the LAME 3.97 alpha stage. This time he compared -V 4 and -V 4 --vbr-new.
http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=31255&view=findpost&p=275554

[lalittle]

It is called progress.

The developers have tweaked the new mode for the last several years.

My point is simply that "fast" is a very poor name for the switch if it also raises quality.  "New" makes a lot more sense.  There is nothing about the word "fast" that implies better quality -- in fact, it tends to imply the opposite if you're not specifically familiar with this.

Note that the "fast mode" (i.e. -f) was originally described this way:

This switch forces the encoder to use a faster encoding mode, but with a lower quality.

This is why the naming of the "fast" switch has me confused.

This quote is from HA's LAME wiki:

The --vbr-new switch enables the new VBR mode. Lame will encode much faster than the old/default VBR mode. In terms of quality, --vbr-new appears to be better than the old model, but reports of artifacts when using the new model do exist. Despite these possible issues, --vbr-new is currently recommended over the default VBR mode due to both the speed and quality increases afforded by the new algorithm.

So when they say "reports of artifacts... do exist" are they making any comparison to the original algorithm -- i.e. are they saying that there is a possibility of artifacts OTHER than what you'd get with the original algotrithm, or are they simply saying that it's not entirely transparent (which is the same thing you could say about the original)?

Thanks,

Larry

[lalittle]

This quote is from HA's LAME wiki:

I just noticed this quote from MC itself.  This is what it says next to fast mode checkbox in the encoding options:

Use Fast Mode (much faster but slightly lower quality)

You can see where the confusion come from.

Is this just a holdover from when "fast mode" wasn't as good?

Thanks,

Larry

[Alex B]

I just noticed this quote from MC itself.  This is what it says next to fast mode checkbox in the encoding options:
"Use Fast Mode (much faster but slightly lower quality)"

You can see where the confusion come from.

Is this just a holdover from when "fast mode" wasn't as good?

Exactly

A quote from my first post:

The Fast switch could be removed and the --vbr-new mode could be the default mode. At least the now incorrect "much faster, but slightly lower quality" text should be removed.

Those who want to tweak LAME options can continue using the custom text box.

When the current options were introduced in MC11.0 JRiver updated them to the LAME recommendations (before that the preset system was not used by default.). I just think that the default options should be updated again to the recommended LAME settings and especially the portable switches would be nice to have.

The LAME --vbr-new mode has advanced especially in the middle range (V5 and so). The highest VBR settings have always been good enough for most users. "--vbr-new or not" does not make a significant difference at V2 - V0. Many MC users are happy with their old MP3 "VBR High" encodings that are from time when the LAME presets were not used in MC by default and LAME's standard VBR mode was not internally mapped with tweaked presets like it is now.

I have a lot of files that I encoded with LAME 3.90-3.92 at -V 1 without any other switches many years ago (before I found MJ and information about presets in HA). Most of the files sound fine and I have not bothered to replace them even I have re-ripped my CDs in APE format for my archive.

If someone prefers to use the VBR mode without the Fast switch nothing would stop from clicking the custom checkbox and removing the --vbr-new part from the command line. The adjusted command line would be preserved in the options until the next settings change.

If the Fast box will be preserved then the text after it should be changed to one word: "recommended".

[johnnyboy]

So what is the current absolute best quality switch to use in LAME to get the absolute highest quality mp3's not worrying about size at all?

Thanks

[lalittle]

So what is the current absolute best quality switch to use in LAME to get the absolute highest quality mp3's not worrying about size at all?

Thanks

According to the info I see on the web, the highest quality you can get with the LAME encoder is to use the switch:

-b 320

This is constant bitrate (CBR) 320kbps.  The difference between this and the highest VBR setting, however, ("-V 0 --vbr-new" or "--preset fast extreme") would likely be totally undetectable, and would average only around 245kbps.

Larry

[Alex B]

As Larry said, the answer is CBR 320 kbps (LAME switch: -b 320 or --preset cbr 320 or --alt-preset cbr 320 or --preset insane or --alt-preset insane, these are mapped with the same internal setting in the current encoder version).

It's nickname is "insane", but you asked for the best possible setting.

It is insane because it uses space inefficiently by not allowing the encoder to lower the bitrate when 320 kbps is not needed for the same perceived quality. However, known problem samples exist that cause some audible artifacts with all VBR modes, but not with CBR 320 kbps. Some problem samples cause audible artifacts with CBR 320 kbps too.

[lalittle]

A quote from my first post:

With all the forum reading I've been doing lately, I completely forgot you said that.  Thanks for re-pointing that out, and for all the details.

If it's the consensus among the experts that the new algorithm is the recommended setting now, I agree that the text should be changed to reflect this -- which is what you said in the first place.  It's difficult to trust it when it specifically says it's "lower quality."

Thanks,

Larry

[johnnyboy]

Thanks. Was just after clarification

Sounds like -V 0 --vbr-new is the one to use then.

For a non audiophile, if I was using high end equipment (not super high end, like a $300 Oki receiver going to some reasonable speakers) would I be able to hear much noticable difference between the mp3 and the original CD using these settings for say the Mid and high range tones?
Like if I wanted to hear the instruments clearly with seperation?

thanks

[Alex B]

Sounds like -V 0 --vbr-new is the one to use then.

For a non audiophile, if I was using high end equipment (not super high end, like a $300 Oki receiver going to some reasonable speakers) would I be able to hear much noticable difference between the mp3 and the original CD using these settings for say the Mid and high range tones?
Like if I wanted to hear the instruments clearly with seperation?

It should be fine.

Usually if LAME has a problem with certain audio signal at -V0 quality level the error makes something sound a bit different for a short moment. This kind of artifact does not necessarily sound bad.

Any decent hi-fi system may be able to reveal this type of problems, but usually a direct comparison with the lossless source file is needed.

[InflatableMouse]

Are you limited to using Mp3?

IMHO Ogg can safely be used using -q 4 without any audible difference until you move into the "audiophile class" amps and speakers.

Wihtout wanting to open a can of worms, I believe Ogg is quite a few steps better in terms of sound quality compared to mp3 on the same average bitrate.

[Alex B]

Ogg Vorbis aoTuV 4.51b -Q 4.25 has about the same quality with LAME 3.97b2 -V 5 --vbr-new. The above mentioned test proved that. The difference was within the error margin.

Though, I know that Vorbis is clearly better than LAME with one of the tested samples despite the overall test results. How do I know that? Answer: I took part in the test and and tested all 6x18 samples. My results are included in the test results.

Personally, I rip in a lossless format and have used Ogg Vorbis and Musepack when MP3 is not a necessity, but this thread is about LAME.

Edit: On second thought, let's try to keep this on topic.

[benn600]

I find, in my own test, that the new, faster mode, is about 20% faster.  With the slow mode, it was estimating 5.5 hours but now it's estimating 4.2 hours.  With the faster mode, I am only using 2 simultaneous conversions and my CPU is averaging 90% vs. the 100% with 3 simultaneous conversions.  This is independent of the faster mode...it doesn't use all my CPU unless I run 3 simultaneous encodes with either mode.

[benn600]

I entered this into the command line switches box:
-V 5 --vbr-new
I am finding that files are still quite large.  Files range from 4-8MB.  Theoretically, I should get 1,500 files on 6GB based on 4 minute 128Kb recordings.  This VBR mode should average around 130 people said and, assuming my songs average a little longer, then I should still get a decent number of songs.

I just checked and the cover art stored is incredible.  I'm betting the extra baggage is from my 1-2 MB cover art.  That adds 1-2MB PER SONG!  Since it's stored inside the mp3 files.  I have a thread going asking if this cover art can be resized before added to the mp3.

[lalittle]

I just checked and the cover art stored is incredible.  I'm betting the extra baggage is from my 1-2 MB cover art.  That adds 1-2MB PER SONG!  Since it's stored inside the mp3 files.  I have a thread going asking if this cover art can be resized before added to the mp3.

Why are you using such massive cover art files?  1 to 2 megs for a single cover art file seems like absurd overkill for this purpose.

Thanks,

Larry

[benn600]

Well, that's what I scanned everything at and I really don't want to throw away all that data.  Is there a LAME command line option to resize cover art because it appears the cover art is stored inside the MP3 files.  Even if you use smaller cover art at 500KB or even smaller, 100KB, it could be shrunk to 20KB and end up saving you a few hundred MB...or in my case, a GB or two.

I really need a way to get this shrunk down before it is added to the file.

[lalittle]

Well, that's what I scanned everything at and I really don't want to throw away all that data.  Is there a LAME command line option to resize cover art because it appears the cover art is stored inside the MP3 files.  Even if you use smaller cover art at 500KB or even smaller, 100KB, it could be shrunk to 20KB and end up saving you a few hundred MB...or in my case, a GB or two.

I really need a way to get this shrunk down before it is added to the file.

I've never heard of such a command, but even if no such command exists, it's possible that somebody out there created some sort of script that does what you're asking.  You should search the various LAME or mp3 related forums around the web to see about this.

That said, I would recommend using MUCH smaller cover art files from now on.  Most people use art around 50K or smaller, which creates a very satisfactory image at 500 x 500 (which seems to be the current "higher end standard" for cover art, and is used on sites like Amazon and Walmart.)  At 1 to 2 megs per album cover, you're using a resolution/quality that is pretty much unheard of in this application.

Larry

[BartMan01]

Well, that's what I scanned everything at and I really don't want to throw away all that data.  Is there a LAME command line option to resize cover art because it appears the cover art is stored inside the MP3 files.  Even if you use smaller cover art at 500KB or even smaller, 100KB, it could be shrunk to 20KB and end up saving you a few hundred MB...or in my case, a GB or two.

I really need a way to get this shrunk down before it is added to the file.

Certainly keep the original scan, but use a down sized 500x500 version for your 'cover art'.  Most image editing programs can batch this for you.

[MadJewDisaster]

~~~~~~As Larry said, the answer is CBR 320 kbps (LAME switch: -b 320 or --preset cbr 320 or --alt-preset cbr 320 or --preset insane or --alt-preset insane, these are mapped with the same internal setting in the current encoder version).~~~~

NEVER EVER encode in 320 CBR - sound is HORRIBLE , close to a old Xing 128 CBR

A MusicMatch cbr 128 sounds better than a 320

To Alex B
No ABX , just a fair external soundcard , a very fair amp and good speakers- And playing always LOUD
My amp having only 8-9 and 10 as volume level =)

Make me a playlist of mix  formats, with few Xing 128 and few 320 mp3 - make this 2 to be play one after one , like 3 xing 128 in a row-

At 100% i will tell you when Xing 128 or mp3 320.
Very easy , you cannot stand it at loud volume - Like hiss , like if someone sending ultrasounds making your ears suffering and you going mad -- you feel like hammering the speakers.You just stop playing cause it is unbearable.

My brother did me few dvds with mp3 lame 320 on it, itranscoded it to APX-You may lost some quality [ still to be proved] but at least this hiss,droning effect is gone.

[Matt]

Build 165 and later will implement Alex's suggestions.  Thanks Alex!

Alex, I sent you an email with a question.  We've had spam-filter issues in the past, so let me know if you don't receive it.

[Alex B]

Matt, please check your forum PM.

[lalittle]

Currently the Fast mode is generally as good as the old mode and often better. Only a few minor problems have been reported.

It's that "few minor problems" thing that alway has me wondering.  The consensus at HA is that while the "fast" switch IS recommended, it's also reported that it can "sometimes lead to artifacts."  What I'm unclear about is whether or not these artifacts are something that the non-fast encoding can also do, or if this is a problem that is unique to the "fast" switch.  The reports are not clear on this.

In other words, I understand that the fast switch will "usually" sound better, but I'm wondering about introducing artifacts that you do NOT get when NOT using the switch.

Thanks for any further feedback on this,

Larry

[Alex B]

MP3 encoding creates always so called artifacts. Usually these artifacts are too small to be audible when high quality VBR settings are used.

These two encoding modes may create different artifacts with the same audio sample. After various listening tests with known problem samples the consensus is that LAME 3.97 --vbr-new creates less audible artifacts on average than the old mode. In casual listening it is unlikely that you hear any difference between the modes.

However, the new default VBR encoding mode in the build 165 is just a recommendation that is based on the available knowledge. You can use the old mode too. Just select your preferred VBR quality level first and then tick the custom option and remove the --vbr-new part from the command line. MC will remember this setting.

If you are worried about the encoding quality you should try a personal listening test. Here's info about a free blind listening test tool: ABC/Hidden Reference Audio Comparison tool.

Personally, I need to use problem samples to distinguish -V 5 (~130 kbps) from the original in a blind test. Making difference between the new and old VBR modes at higher bitrates is much more difficult.

[benn600]

Back to the high resolution cover art, I have a great idea.  I can store the original scanned 1200x1200 image as a different filename than folder.jpg and folder.jpg will be a resized image.  Now I'll go find a batch resizer.  Then I'll need to batch rename.  Hey, MC could do that if I didn't exclude folder.jpg files during importing.