What is your favorite MP3 encoder?

[OmegaZero]

What I am looking for is the highest quality playback I can get in an MP3 format. I know I really should be using a lossless format, but that is just not an option for me.

I've run into serious quality issues with the VBR implementations used by some of the software packages out there these days. I figure there must be a good one or two, and if so someone here has probably found it.

What do you use?

[frostedflakes]

Have you tried LAME? Among the audiophiles, it's considered to be the highest quality MP3 encoder out there. You can grab it here.

Not sure what issues you were having with VBR, some older MP3 players don't support it, but anything made in the last five years or so should. I use LAME V2 preset and haven't had any problems. Maybe other encoders are just really bad with VBR, definitely give LAME VBR a shot if you haven't already.

[jessekopelman]

I know I really should be using a lossless format, but that is just not an option for me.

Why do you know this? Double blind testing shows that almost nobody can tell the difference between lossless and 256kbps and higher VBR. Lossless is good for archival purposes, but not necessary for fine listening.

Anyway, iTunes produces top notch MP3. Unfortunately it is serious bloatware if you are going to use it just for encoding. LAME is also excellent, but you need a good front end for it (unless you enjoy writing batch files). I used to use Foobar2000 with good success.

[Stravos]

Exact Audio Copy (front end) and LAME (encoder) seems to be the go-to combo for enthusiasts. There are plenty of step-by-step guides out there for setting it all up. EAC has improved it's setup a lot, though, so it's pretty easy to just install and start ripping within minutes now.

[Blacktree]

LAME + EAC (Exact Audio Copy) is considered to be the "dynamic duo" of MP3 encoding. I use this combo, and am very happy with it.

[widowmaker]

Why do you know this? Double blind testing shows that almost nobody can tell the difference between lossless and 256kbps and higher VBR. Lossless is good for archival purposes, but not necessary for fine listening.

I might be in the "nobody" category, but I can definately tell the difference between 320 and flac. You can notice a huge difference simply by using optical media. A pair of good headphones are a must. Listening to lossless music on a computer doesn't sound much different than 320 mp3's (maybe my soundcard just sucks) but if you try burning both the mp3 and lossless onto a CD and listening to it off a CD player, it'll blow your mind. I've since been looking for a portable DVD audio player with no LCD screen. It seems like nobody wants to make it.

[merlin]

Actually even on my friend's 10k sound system, I can't tell 320k from lossless. Anyways, Lame + EAC is awesome, I've used it on 200+ cds and it's great. Only minus is no unicode, but get a tagger to fix any unicode cds.

[OmegaZero]

I mentioned lossless in my original post in an effort to keep the thread focused. I figured if I didn't then half of the replies would be about the differences between lossless and MP3. Clearly it worked...

Anyway, thank you all for you recommendations. I am working with EAC now and it seems to be good. The only real hiccup has been the fact that it seems to incorrectly pass compression settings to LAME. I am working on writing the command-line switches manually - should fix the issue. Has anyone else had this issue? If so, I would love to see your command-line settings for comparison.

PS: In case anyone was wondering - my parameters so far :
-m s -q 0 -v --vbr-new -V 0 -b 112 --replaygain-accurate --clipdetect --tc "EAC 0.99pb4 & LAME 3.97" --tn %n --add-id3v2 %s %d

[Ant6n]

...
Listening to lossless music on a computer doesn't sound much different than 320 mp3's (maybe my soundcard just sucks) but if you try burning both the mp3 and lossless onto a CD and listening to it off a CD player, it'll blow your mind. I've since been looking for a portable DVD audio player with no LCD screen. It seems like nobody wants to make it.

maybe the decoders suck

[blackworx]

...
Listening to lossless music on a computer doesn't sound much different than 320 mp3's (maybe my soundcard just sucks) but if you try burning both the mp3 and lossless onto a CD and listening to it off a CD player, it'll blow your mind. I've since been looking for a portable DVD audio player with no LCD screen. It seems like nobody wants to make it.

maybe the decoders suck

Much more likely to be the audio hardware. If the analogue stage is inside your PC, then no matter how good your sound card is it will be subjected to a lot of RF/EM interference. Hence why with crummy onboard audio one can often hear mouse movement translated into audible noise on your speakers/headphones etc. (Apologies if I'm preaching to the choir.) Coaxial SPDIF + external DAC ftw imho.

Anyway, in response to op: +1 vote for the EAC+LAME combo.

Ed: You might also want to check the DAE capabilities of your source drive. This site is useful: http://www.daefeatures.co.uk/search.php

[nick705]

The only real hiccup has been the fact that it seems to incorrectly pass compression settings to LAME. I am working on writing the command-line switches manually - should fix the issue. Has anyone else had this issue? If so, I would love to see your command-line settings for comparison.

PS: In case anyone was wondering - my parameters so far :
-m s -q 0 -v --vbr-new -V 0 -b 112 --replaygain-accurate --clipdetect --tc "EAC 0.99pb4 & LAME 3.97" --tn %n --add-id3v2 %s %d

I haven't ripped any CDs to MP3 for ages, but I don't remember this happening...

LAME's default settings are pretty well tuned these days BTW, and overriding them with command line parameters usually does more harm than good... for instance, -m s disables joint stereo, which bloats the file size and/or means less bits will be available in other places where they're really needed. Specifying a lower bitrate limit with -b 112 is unnecessary as well - it may have been worthwhile a long time ago, when the encoder was relatively immature and couldn't be trusted not to produce artefacts when left to its own devices, but it's probably pointless these days.

The HydrogenAudio wiki is a good starting point for more LAME info if you're interested.

Tempting as it is, I won't get into the argument about 320kbps LAME being easily distinguishable from the source material, as it always seems to end up in a bad place... suffice it to say, any listening tests that aren't performed under double-blind ABX conditions are hopelessly compromised by the placebo effect and are effectively worthless.

[OmegaZero]

nick705: I found that despite the settings I specified within the EAC GUI, when I look at the LAME window it still said VBR_q=2 and old VBR were being used. Not sure why that happened, but it was easy enough to get around (with the help of a little CLI documnetation).

-m s disables joint stereo, which bloats the file size and/or means less bits will be available in other places where they're really needed.

According to the LAME usage guide, straight stereo is going to be the default for bitrates above 128 and VBR_q <=4. That said, it is probably pointless to specify but I figured since I was writing the CLI options anyway I might as well specify everything. Also, I did test the file sizes with and without this setting and found them to be more or less identical.

I've never heard or read that "less bits will be available in other places" due to stereo settings before. Do you have a source? I would imagine that if that were true stereo would not be the default mode for high quality settings...

Specifying a lower bitrate limit with -b 112 is unnecessary as well - it may have been worthwhile a long time ago, when the encoder was relatively immature and couldn't be trusted not to produce artefacts when left to its own devices, but it's probably pointless these days.

This was also a recommended setting from the LAME usage guide (current version). According to that doc it protects against an intermittnat issue with VBR where the compression may get too aggressive and cause distortion regardless of how high the quality settings are.



Command-line specification aside, I am pretty happy with EAC. Thank you all for your input!

[nick705]

According to the LAME usage guide, straight stereo is going to be the default for bitrates above 128 and VBR_q <=4. That said, it is probably pointless to specify but I figured since I was writing the CLI options anyway I might as well specify everything. Also, I did test the file sizes with and without this setting and found them to be more or less identical.

I've never heard or read that "less bits will be available in other places" due to stereo settings before. Do you have a source? I would imagine that if that were true stereo would not be the default mode for high quality settings...

What I was getting at is that you might have a complex passage that doesn't compress very well, but has little stereo separation - in that case, with a limited number of kbps to play with, there's no point wasting bits encoding each channel seperately when it provides no audible benefit, and those bits could be put to better use providing extra headroom where it's needed (this would admittedly help more at lower settings and fixed bitrates). The LAME documentation itself suggests as much, and people are regularly roasted alive on the HA forums for using the -m s switch (do a search there for "joint stereo" and you'll see what I mean).

You're quite right, though, the LAME documentation also suggests that joint stereo isn't used anyway for the higher settings, which is something I hadn't realised. I'm not sure it's entirely accurate though, or maybe it means something different from what it initially suggests - if you run LAME from the command line, even with the -V0 parameter, the output suggests that joint stereo is still being used, like so:

LAME 3.97 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz
Encoding 01 - In My Secret Life.wav to 01 - In My Secret Life.mp3
Encoding as 44.1 kHz VBR(q=0) j-stereo MPEG-1 Layer III (ca. 5.7x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
11328/11328 (100%)| 0:15/ 0:15| 0:15/ 0:15| 19.096x| 0:00
32 [ 105] %*
40 [ 1] %
48 [ 1] %
56 [ 2] %
64 [ 4] %
80 [ 4] %
96 [ 14] %
112 [ 8] %
128 [ 9] %
160 [ 149] %*
192 [ 1580] %%%%%%**************
224 [ 5483] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*************************
256 [ 3006] %%%%%%%%%%%%%%%%%%%%%****************
320 [ 962] %%%%%%%*****
-------------------------------------------------------------------------------
kbps LR MS % long switch short %
233.1 54.4 45.6 92.0 4.7 3.3
Writing LAME Tag...done
ReplayGain: -3.8dB


I just used the "-V0" switch without any other arguments - the "%" symbols represent left-right (full stereo) frames, the "*" symbols are mid-side (joint stereo).

Using "-V2" for the same track, joint stereo is used even more heavily, in fact only 14.3% of the frames are encoded in full stereo:


LAME 3.97 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 01 - In My Secret Life.wav to 01 - In My Secret Life.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
11328/11328 (100%)| 0:17/ 0:17| 0:17/ 0:17| 17.140x| 0:00
32 [ 113] %*
40 [ 1] *
48 [ 8] *
56 [ 10] *
64 [ 4] *
80 [ 6] *
96 [ 26] *
112 [ 195] ***
128 [ 542] %*******
160 [ 3769] %%***************************************************
192 [ 4794] %%%%%%%%%%%%%%%****************************************************
224 [ 1401] %%%%%%**************
256 [ 371] %*****
320 [ 88] %*
-------------------------------------------------------------------------------
kbps LR MS % long switch short %
181.8 14.3 85.7 92.6 4.4 2.9
Writing LAME Tag...done
ReplayGain: -3.8dB

Here it is again using -V2 -m s:

LAME 3.97 32bits (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE (ASM used), SSE2
Using polyphase lowpass filter, transition band: 18671 Hz - 19205 Hz
Encoding 01 - In My Secret Life.wav to 01 - In My Secret Life.mp3
Encoding as 44.1 kHz VBR(q=2) stereo MPEG-1 Layer III (ca. 7.3x) qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
11328/11328 (100%)| 0:15/ 0:15| 0:15/ 0:15| 19.565x| 0:00
32 [ 112] %%
40 [ 4] %
48 [ 8] %
56 [ 5] %
64 [ 6] %
80 [ 4] %
96 [ 2] %
112 [ 2] %
128 [ 6] %
160 [ 1319] %%%%%%%%%%%%%%%%
192 [ 5531] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
224 [ 2823] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
256 [ 1092] %%%%%%%%%%%%%%
320 [ 414] %%%%%%
-------------------------------------------------------------------------------
kbps LR % long switch short %
205.1 100.0 92.8 4.3 2.8
Writing LAME Tag...done
ReplayGain: -3.8dB

All the frames are full stereo, as you'd expect, and the overall bitrate has gone up from 181.8 to 205.1 kbps. For a CBR setting, there would obviously be no leeway to flex upwards where necessary, so quality might suffer if there weren't enough bits available when needed for other purposes.

I guess after all that, the best recommendation would be to trust your own ears (having done the appropriate ABX testing, naturally)...

[jbw]

I used EAC + LAME exclusively until I also heard good things about CDEX. Turns out that CDEX + LAME works just as well for non-scratched CDs and it's much faster (at least on my drive).

[wim]

a few years ago i ripped my entire cd collection using LAME (--alt-preset standard, at the time, seems to be same as -V 2 now) and EAC frontend. it was painless and the results indistinguishable from cd quality for me, highly recommended

[Avalanche]

Turns out that CDEX + LAME works just as well for non-scratched CDs and it's much faster (at least on my drive)."

That echos my experience completely. On my laptop, for some reason, EAC seems to want to read the CD several times over, even brand new ones.

[Meowbay]

I might be in the "nobody" category, but I can definately tell the difference between 320 and flac.

Sorry but that is simply a lie.
It has been proven time and again, using blind tests, that even high-end listeners, classical music freaks, cannot pick the original when using LAME 3.97/3.98 starting around ~224 kbit/s VBR for your MP3 compression.

I highly doubt you will pass the test guessing what the original and what the mp3 are. You will not score more than 50% correct. Trust me, you just will not, because you can't tell the difference.

I recommend you to read these:
ftp://ftp.jult.nl/pub/ten_biggest_lies_in_audio.pdf
http://jthz.com/mp3/#MYTH

[nick705]

Sorry but that is simply a lie.

I'm sceptical as well, but many people do genuinely believe they can always hear a difference between a lossy compressed file and the source.

However dubious their claims, calling someone a liar without a very good reason is inflammatory, rude and is unlikely to take the argument in any constructive direction.

[Vicotnik]

I have tested myself and can definitely spot a 128kpbs MP3, but at 192kbps I cannot tell the difference from the source.

I rip all my CDs using EAC + FLAC. FLAC because it's lossless and HDD space is cheap. When I decode my FLACs and recode the resulting WAVs to MP3s for mobile use, I use Lame with the --preset standard setting.

For listening to music well coded MP3s are ok, but I would never archive my music using a lossy format. Not today anyway, when HDD space is so affordable.

So when I listen to music at home it's FLAC, when I'm on the move with my portable player, it's MP3.

I might be in the "nobody" category, but I can definately tell the difference between 320 and flac. You can notice a huge difference simply by using optical media.

That is possible but I really doubt it. If you indeed hear the difference you are part of the tiny, tiny, tiny minority that do. The chance of that is so small that I don't hessitate to assume that the test you took wasn't a double blind test (or even a single blind).

Even if you can hear the difference it's pretty much proven that that is not the case with the wast majority of people. So suggesting using optical media is pretty pointless. It's been done. All serious double blind tests comes to the same conclusion.

You will not score more than 50% correct. Trust me, you just will not, because you can't tell the difference.

He will probably score around 50%. But telling him that he will not score more than 50% is like telling someone that there is no way he can flip heads two times in a row.

And like nick705 points out, calling widowmaker a liar is very rude. He is most likely misstaken but that doesn't make him a liar.

[widowmaker]

All I'm saying is that I can hear the difference between 320 and lossless when you copy them onto cd's and play them off a sound system. I however cannot hear a difference between them on my computer through a Sound Blaster X-FI Xtreme Audio.

[wim]

it is significant here to distinguish between believing you can hear difference and demonstrating you can hear difference.
even if someone is unable to demonstrate they can tell lossy codec and source pcm apart, if they believe (or imagine) they can hear some difference then that may detriment from their listening experience. and the goal at the end of it all is to have a positive listening experience. the human psychology is a complicated and illogical thing
so let everybody use whatever bitrate and/or format makes them happy

personally i'm happy with vanilla LAME at basically any bitrate above 160k, because considering the only thing i use music for (i.e. listening) there would be no benefit and indeed no detectable difference to any higher rate.

if i felt that some measure of quality-per-byte was not as important as absolute quality, then the only logical conclusion would be to stop listening to recorded music altogether 44.1kHz PCM included. no amount of sophisticated technology can exactly reproduce the sound of, say, plucking an acoustic guitar string by using a some cones and magnets ..

[Vicotnik]

All I'm saying is that I can hear the difference between 320 and lossless when you copy them onto cd's and play them off a sound system. I however cannot hear a difference between them on my computer through a Sound Blaster X-FI Xtreme Audio.

But how can you be sure? Have you tested it double-blind? Even if you have, the chance of something being wrong with your testing method is greater than the chance of you actually hearing a difference.

What you claim is so unlikely I would almost have to see it to believe it. There is simply so much tests out there that claims the opposite of what you say that a simple "I can hear the difference"-statement doesn't say much.

it is significant here to distinguish between believing you can hear difference and demonstrating you can hear difference.
even if someone is unable to demonstrate they can tell lossy codec and source pcm apart, if they believe (or imagine) they can hear some difference then that may detriment from their listening experience. and the goal at the end of it all is to have a positive listening experience. the human psychology is a complicated and illogical thing
so let everybody use whatever bitrate and/or format makes them happy

True. But there is nothing wrong with informing people of "the scientific method". I like listening to FLACs because then I know there are no compression artifacts. Even if the risk of me hearing an artifact in a well coded MP3 is virtually non-existent.
But I don't think anyone is better of believing they can hear a difference for real-real, and that most others can as well.