Hi people, first time here and a have a couple of questions that i hope someone can answer me:

i looked around for a bit in the forum and didn't see a topic about this specifically (if there is i apologize): what are the best settings to use with cdex and LAME (i usually use VBR mode, i think it's better) in order to get the best compromise between the sound quality of the mp3 and the file size?
because i see that there is a topic about the settings for best quality but i would like to know the settings for the best compromise, i want to backup all my music cds to the computer with superior mp3 quality but not having to encode everything at 320kbps!

also, i've read stuff on the net about joint stereo vs stereo and there does not seem to be a consensus about joint stereo being lossy or not. is it?

i use cdex 1.51 and LAME 3.97

thanks is advance!

Read the wiki page if you haven't already:

http://wiki.hydrogenaudio.org/index.php?title=LAME

If you're not a stickler for transparent audio, try -V6 --vbr-new. Anything beyond that resamples the audio which can cause noticeable degradation and/or compatibility problems with setups expecting only 44100 or 48000 hz audio. Of course, you can also start from -V9 --vbr-new and work your way up until you are satisfied with results, though conventional wisdom is that ABR is better for modes below -V6 --vbr-new. I wouldn't know since I don't use MP3 with low bitrates.

EDIT: Add --vbr-new

ok, thx for the reply,

for what i gather i definetely am a stickler for transparent audio.

i read the wiki page but basically what is there is what is here in the forum about the best vbr settings possible... i should then use one of those mentioned settings for what i intend, right? for the best compromise?

what about that stereo vs joint stereo thing?

"The model-human does not exist - at all!" - Lyx

What should that mean? Well, it means that the efficiency-optimum is different for every individual, because the hearing, room-acoustics, equipment, music, listening-style and so on is different for every human.

For "most people", the quality/filesize optimum appears to be at around V5, take or give one.

You need no additional switches.

- Lyx

Not even --vbr-new?

The interconversion between left/right and mid/side is essentially lossless. It's the encoding of either L/R or M/S to MP3 that's lossy. Whether L/R or M/S gives the best sound when encoded varies from one frame to the next. That's why the encoder (if well designed and implemented) makes that decision while it is encoding. If you trust the encoder to make the right choice (and LAME reportedly does) then trust it to choose for you by using joint stereo.

Yes, use --vbr-new, as this is recommended for LAME 3.97, the current recommended compile in the Wiki. In LAME 3.98, --vbr-new will become the default, so can then be retained or omitted with no effect.

I'd agree that -V5 --vbr-new is considered to be good quality and small size. Most of the differences noticed under blind testing or original versus -V5 --vbr-new seem to rate as roughly score=4 (perceptible but not annoying).

-V2 --vbr-new was designed to be indistinguishable most of the time under pretty demanding listening conditions (we call this "transparency"), and is essentially the new improved version for what was "--preset standard". It does get forced to waste bits to retain quality because of an inadequacy in the MP3 specification for the highest frequency band, sfb21.

-V3 --vbr-new is also worth a look because it's almost as finely tuned for transparency as -V2, but it over-rides the bit-wasting workaround for the sfb21 problem to occasionally allow more distortion in the top-frequency band which is barely audible in most music anyway. This allows it to save a good deal of bitrate for very little loss in quality.

Personally, as I ALWAYS listen in Album Gain mode (except for a few iTunes+ tracks for which I only have Track Gain) so I've started to scan for Album Gain first (often from a lossless EAC rip) and then encode using foobar2000 with Album Gain pre-applied, outputting the loudness-adjusted audio in 24-bit straight to LAME. This often (not always) tends to reduce the bitrate bloat with -V2 --vbr-new though I'm not sure about whether it reduces bitrate at all with -V3 or -V5 at all (it might do slightly, given the ATH model in LAME).

I've been back and forth but I am done with high bitrate encodings. The sense of satisfaction was always short lived. I found -V4 to be the best tradeoff. Quality at 130k is quite good today, so -V4 gives me extra headroom and still very small files that don't need to be transcoded.

I never found too much improvement over -V4 - If it sounds bad then -v2 isn't too good either..

Thank you for explaining that behavior.


In "casual" listening, I generally can't tell the difference until the V5-V6 range. When I do find an issue, I, too, find it generally carries throughout quality settings. I have the same problem with Vorbis; if it's offensive in -q2, it'll probably be offensive at -q5, too.

A while back I sat down and decided to make some comparisons and record the results. It was using Lame 3.92, and I compared 192 CBR, 192 ABR, 224 ABR, APS and APE. I ripped the same track with the same software and just kept trying the different settings.

In terms of size, the track ended up varying from 4214 to 5135, average of 4635.

Going from APS to APE added about 1 MB in size (4281 to 5114) and the bitrate went up about 40k, from an average of 194 to an average of 232. And to be honest, I'm not sure I could tell the difference. I've been using APS ever since.

I suggest you do something similar and then try and let your ears be the judge of what you really need.

well, I've got sensitive ears and good equipment, so I rather stick to -V1 --vbr-new, same lowpass as with -V2 but few more bits (which does not make me worrying when I look at HDD prices) cause me the warmfuzzyfeelingTM behind my a*s anyway, I am a practical person who does not like reripping collection in future just because of few wasted bits in present time..