MP3 questions

For most people, the final medium for their songs is 44.1/16-bit on a CD. But for people like me, who mostly just distribute their songs on the internet, the final format is MP3. In this case, 128 KBPS.

My question is, is there anything you can do during tracking/mixing/mastering to ensure you get a good sounding MP3?

Should I first make a it sound great for CD? And then make a separate master just for MP3s? What would I do differently?

I'm perfectly happy with the way things sound on CD but when converting to MP3 they sound like shit. Keep in mind I mainly do acoustic guitar so that might make it more obvious.

One thing I learned is to use very little, if any, reverb as that sounds like ass at 128 KBPS. So for my CD version I'll keep a little reverb, but for the MP3 version I'll usually remove it completely.

Also, what frequencies are usually most effected when converting to 128 KBPS? Should I maybe try lowering those nasty frequencies a bit and boost the ones that are less effected to give a better sound?

If you run a professional recording/mastering studio and your client also wants a good sounding MP3 what steps would you take? Just convert to MP3 and be done?

[NTK88]

IDK for sure but I find my cymbals sound like shit when I save 128bit mp3.
Probably high end freqs, just the ones to make acoustic guitar shine!!!

[MadAudio]

I can't convert to anything less than 192 because it just sounds like shit to me.

Originally Posted by NTK88

IDK for sure but I find my cymbals sound like shit when I save 128bit mp3.
Probably high end freqs, just the ones to make acoustic guitar shine!!!

Yup, go figure. The low-end/mid range always sounds pretty good; bass guitar, vocals, etc. Just what most of my tunes don't have.

Originally Posted by MadAudio

I can't convert to anything less than 192 because it just sounds like shit to me.

Agreed. When I have the option I always use 192. If it's a really short tune I'll use 320. If only the majority of song sites used 192 instead of 128, but that would be a lot of increased bandwidth.

[MadAudio]

Originally Posted by danny.guitar

Agreed. When I have the option I always use 192. If it's a really short tune I'll use 320. If only the majority of song sites used 192 instead of 128, but that would be a lot of increased bandwidth.

That's what I like about lightningmp3.com - the only restriction is on file size. As long as it's under 16 megs, you can encode it however you like.

[talontsiawd]

I don't have alot to add but i'll say this. The obvious, the better the wav file sounds, the better the mp3 sounds.

Second, most people i know who constantly listen to mp3's and don't make music really will claim they don't hear much difference between a wave file and a 128kps mp3. And i'm inclinded to believe it's alot easier to tell when things sound worse than better. So if your audience is used to that, the quality loss can be concider negligable i guess.

Now, the main thing i have noticed with online media players is they do sound different. Myspace for example sounds really flat and dull compared to say windows media player. So trying to mix to that format might be counter productive, if there is a good way to do it to begin with.

Originally Posted by talontsiawd

I don't have alot to add but i'll say this. The obvious, the better the wav file sounds, the better the mp3 sounds.

Yea, but I'm already happy with the way things sound as WAV. And since my songs usually are just solo acoustic guitar, it's much more obvious in MP3 format.

Originally Posted by talontsiawd

Second, most people i know who constantly listen to mp3's and don't make music really will claim they don't hear much difference between a wave file and a 128kps mp3. And i'm inclinded to believe it's alot easier to tell when things sound worse than better. So if your audience is used to that, the quality loss can be concider negligable i guess.

That's a good point. But I do have a lot of people that listen that are also musicians and they've said they can hear a night & day difference in quality. Much more with my songs than with other ones (probably because it's just acoustic guitar).

Originally Posted by talontsiawd

Now, the main thing i have noticed with online media players is they do sound different. Myspace for example sounds really flat and dull compared to say windows media player. So trying to mix to that format might be counter productive, if there is a good way to do it to begin with.

I think the reason MySpace sounds so bad is because the MP3s are 96KBPS, which is even worse than 128. So that's one reason I don't put any songs on MySpace anymore.

[travelin travis]

Originally Posted by MadAudio

I can't convert to anything less than 192 because it just sounds like shit to me.

Now if only myspace would think that way.

[LogicDeLuxe]

Originally Posted by danny.guitar

Should I first make a it sound great for CD? And then make a separate master just for MP3s? What would I do differently?

There is no need to encode from a CD master. You can encode to MP3 directly from 24 bit or float wav and that way avoid noise shaping which otherwise would eat bitrate. Also clipping (worse enough on a CD) is a bad thing, since it produces harmonics which also take additional bitrate. Better make the MP3 slightly softer, ie. peaking at about -1 dB. Also use a decent encoder like LAME. If you can, use VBR. And ABR 128, of course, is better than CBR 128.
If your reverb was a stereo effect, it sure takes bitrate as well. Mono effects should be less hungry, providing you're encoding joint stereo, which you really should do at low bitrates.
If you used a mono mic, 128 kbps actually should do quite well (unless stereo effects are applied).

[asulger]

Originally Posted by LogicDeLuxe

If you used a mono mic, 128 kbps actually should do quite well (unless stereo effects are applied).

If danny.guitar is talking about soundclick posting, which is limited to 128 for free accounts - I think this is bingo!
I'm going to try it myself - my stuff is all acoustic guitar.
Thanks,

[NL5]

I dunno, but I think MP3 conversion gets blamed for more damage than it really does. Sure, I can hear the difference between 128 MP3 and a CD, but it doesn't totally trash the sound as some people would say. (and I listen in a pretty accurate environment)

Here is a song I am just about done with that I used LAME to convert from a 16/44 mixdown. Pretty clear I'd say. Not as good as the CD version, but definetly not trashed. And, there is stereo 'verb all over it, as well as a bunch of actual room sound.

http://www.lightningmp3.com/live/file.php?fid=6590

.

[Ford Van]

Originally Posted by danny.guitar

Yea, but I'm already happy with the way things sound as WAV. And since my songs usually are just solo acoustic guitar, it's much more obvious in MP3 format.

I listened to your soundclick song. What I noticed is that you seem to have a somewhat shoddy mic on your acoustic. A somewhat bloaty tone on the instrument, with some string cheese stuff happening (fret buzz) that is inappropriate to the style you are playing.

I can see why with these combination of things how encoding to mp3 really messes things up.

Here is a short 20 second sample of a mastered .wav file of something I mixes, along with a Fraunhofer encoded 192kbs and 128kbs mp3's of the same sample. This is acoustic guitar and vocal.

http://www.phoenixlightandsound.com/...test/GEWav.wav
http://www.phoenixlightandsound.com/...test/GE192.mp3
http://www.phoenixlightandsound.com/...test/GE128.mp3

Yeah, there is a slight difference between the .wav and 128kbs file, but not much!

You should check your encoder to make sure you are not encoding as something like a "fast" encode. I use Audio Active Production Studio Pro app, and it gives me the option of a "Faster encode" or a "High Quality Encode" which is much slower, but produces a much better sounding mp3 too!

Here is a 128kbs on "Faster encode":

http://www.phoenixlightandsound.com/.../GEfast128.mp3

Still, not a night and day difference, but certainly you can start to hear artifacts.

I do not like the Lame codec at all! I will not use it. While it is a subtle difference between it and the Fraunhofer codec, it is a difference I don't like. But, I don't think the codec itself is the culprit with the big difference you are hearing. I think the encoding is showing the weaknesses of your mix.

I am always concerned when somebody makes a statement that they are "happy" with their current production results. I have never recorded/mixed/mastered something where I was "satisfied". I just get to a point where it has to be done is all! But I can always find things I would like to improve, thus, I improve every time I work on something.

[NL5]

Originally Posted by Ford Van

I am always concerned when somebody makes a statement that they are "happy" with their current production results. I have never recorded/mixed/mastered something where I was "satisfied". I just get to a point where it has to be done is all! But I can always find things I would like to improve, thus, I improve every time I work on something.

So true - not that I am very good anyway, but I'm that way with everything. Never satisfied, but eventually, I just decide that it's time to call "it" finished - whatever "it" happens to be.......

edit - and WTF is so bad about LAME - I thought it used the Fraunhaufer codec???

[Ford Van]

I can hear the results of your new console for sure on that mix! I should get my butt over there to check it all out. There is a certain something that a decent analog console does to the mix that digital just can't!

I am sure that band was VERY pleased with that mix!

[NL5]

Originally Posted by Ford Van

I can hear the results of your new console for sure on that mix! I should get my butt over there to check it all out. There is a certain something that a decent analog console does to the mix that digital just can't!

I am sure that band was VERY pleased with that mix!

pssssst - don't tell anyone, but that is 100% ITB mix - the console appears NOWHERE on that song. It is up and running, but not quite ready for primetime, and the band kept making changes, so I just kept it ITB to make total recall possible. The Trident was used to monitor during overdubbing, that's it. It really is a shame, cause it sounds badass!

Yep, you need to come check out the new digs - it's a whole new ballgame!

[Ford Van]

Originally Posted by NL5

pssssst - don't tell anyone, but that is 100% ITB mix - the console appears NOWHERE on that song. It is up and running, but not quite ready for primetime, and the band kept making changes, so I just kept it ITB to make total recall possible. The Trident was used to monitor during overdubbing, that's it. It really is a shame, cause it sounds badass!

Yep, you need to come check out the new digs - it's a whole new ballgame!

Interesting.

NL5 - Thanks for posting that clip, sounds good to me. But what I was saying is that for something that is just acoustic guitar, (no vocals, electric guitar, bass, etc.) that maybe the artifacts are more noticeable.

Ford Van - That is an old recording before I got my guitar repaired, it had a lot of string buzz. Some of it was also poor technique because that song is hard as hell to play.

I think I found the problem. I was using the MP3 Pro feature in Adobe Audition because it says it uses the Fraunhofer codec so I thought it would be better.

I then went back to cDex and used LAME and it sounds fine. I honestly don't hear that much of a difference between the WAV and the MP3. It's there, but not like before.

Please tell me if you hear any noticeable artifacts:
www.getmyim.com/temp/guitar_send128.mp3 (~ 800 KB)

[LogicDeLuxe]

Originally Posted by NL5

Here is a song I am just about done with that I used LAME to convert from a 16/44 mixdown. Pretty clear I'd say. Not as good as the CD version, but definetly not trashed. And, there is stereo 'verb all over it, as well as a bunch of actual room sound.

Sure, trashed is a bit hard a judgment. Though there are many download shops selling that quality for almost the same price the CD costs. There should be definately more quality to be interesting for that much money. And WMA sounds even worse at 128 kbps, imho. Fortunately, you can buy a used CD much cheaper in most cases, which render those download shops rather useless.
Another problem is, that not all download shops actually use LAME but rather obsolete encoders, often because they bought them years ago with watermarking algorithms included.

Originally Posted by NL5

and WTF is so bad about LAME - I thought it used the Fraunhaufer codec???

It does not. It's open source developed from scatch. There's nothing bad about it, imho. If the results are worse than with Fraunhofer, you probably gave it an inappropriate parameter.

Originally Posted by danny.guitar

I think I found the problem. I was using the MP3 Pro feature in Adobe Audition because it says it uses the Fraunhofer codec so I thought it would be better.

It has better high frequencies, but they are much less accurate. This works well for percussive instruments, but with some instruments it's probably a bad idea. Also the Player has to sopport MP3pro, if it doesn't your frequency band gets reduced to half.

Please tell me if you hear any noticeable artifacts

Sounds fine to me.
Just a thing I noticed (in the editor), your recording is mono, but the noise floor is stereo, hence wasting a tiny bit space in the mp3. You can try encoding a mono wave, which might improve the result slightly.

Originally Posted by LogicDeLuxe

Just a thing I noticed (in the editor), your recording is mono, but the noise floor is stereo, hence wasting a tiny bit space in the mp3. You can try encoding a mono wave, which might improve the result slightly.

Thanks. The original WAV file was recorded in mono. I converted to stereo afterwards. Otherwise, it would only come out of my left speaker and I thought it would be hard for people to tell how it sounds that way.

[XLR]

What I've noticed with my own conversions from wav to mp3 is that ANY imbalance in the sound gets magnified during the conversion. If it was a little too bassy it gets WAY too bassy. If the highs were just a little harsh they get unbearable. If there was a fingering buzz, it's like a magnifying glass is held over every spot where that happened.

With 320 kbps mp3s it's less so. With 128 it's much more a problem.

The Lame encoder is the best I've found. It makes a difference.

Originally Posted by Timothy Lawler

What I've noticed with my own conversions from wav to mp3 is that ANY imbalance in the sound gets magnified during the conversion. If it was a little too bassy it gets WAY too bassy. If the highs were just a little harsh they get unbearable. If there was a fingering buzz, it's like a magnifying glass is held over every spot where that happened.

With 320 kbps mp3s it's less so. With 128 it's much more a problem.

The Lame encoder is the best I've found. It makes a difference.

Yeah I noticed that too.

Although even at 320 KBPs, or WAV for that matter, my mistakes are still pretty obvious.

[Ford Van]

Originally Posted by danny.guitar

NL5
I think I found the problem. I was using the MP3 Pro feature in Adobe Audition because it says it uses the Fraunhofer codec so I thought it would be better.

I then went back to cDex and used LAME and it sounds fine. I honestly don't hear that much of a difference between the WAV and the MP3. It's there, but not like before.

Please tell me if you hear any noticeable artifacts:
www.getmyim.com/temp/guitar_send128.mp3 (~ 800 KB)

The difference between LAME and Frauhofer codecs in themselves is not going to be very much of a noticeable difference.

The difference is probably in HOW each program does the encoding. AA is possibly doing a "quick" encode, which cDex is doing a slower, more accurate encode.

[LogicDeLuxe]

Originally Posted by danny.guitar

Thanks. The original WAV file was recorded in mono. I converted to stereo afterwards.

I supposed, you also downconverted from 24 bit to 16 bit, right? Doing this, you probably used some kind of dither (applied in stereo), which is generally a good thing for CD masters. If the target is MP3, there is no need to convert to 16 bit.
Also every mp3 player I'm aware of will play mono files on both channels.
But if you really must convert a 24 bit mono file to 16 bit stereo for some reason, go first to 16 bit mono and after that convert it to stereo. That way, you'll get mono dither.

Originally Posted by Timothy Lawler

What I've noticed with my own conversions from wav to mp3 is that ANY imbalance in the sound gets magnified during the conversion. If it was a little too bassy it gets WAY too bassy. If the highs were just a little harsh they get unbearable.

Did you play them in Winamp by any chance? I once noticed, the built in EQ behaves somewhat differently between MP3 and WAV. Also the other way around, when you convert a MP3 to WAV and compare them, as they really should sound the same then. When I disable the EQ, those differences vanish as well.

Originally Posted by Ford Van

AA is possibly doing a "quick" encode, which cDex is doing a slower, more accurate encode.

You can set all the parameters the Fraunhofer codec has to offer in AA. You sure can do it quick if you really want for some reason. Of course, you wouldn't expect the best quality then. The same goes for LAME.

LogicDeluxe - Thanks for the tip, I never thought about the mono vs stereo dither or converting to MP3 straight from the 24-bit file.

Ford Van - I don't see any option for "fast conversion" or whatever in AA.

In cDex there is "On the fly MP3-encoding". Is that what that is? It is checked by default.

Thanks for all the help.

[LogicDeLuxe]

Originally Posted by danny.guitar

I don't see any option for "fast conversion" or whatever in AA.

It should look something like this:

Using "best quality" should be a good idea, apparently.