Mastering / Mixing for MP3, any good tricks?

[Studio de' Lara]

I just started doing alot of MP3's for people. I am losing alot in the MP3 vs. the uncompressed wave. Any thoughts, as to what is the best bit rate, compression type or mix ideas to compensate for the loss in quality, is appreciated.

I have heard some good MP3s, I am just coming up a little short....

Regards,
Rich

[Bob L]

I use 192 bitrate and straight stereo format... my mp3's sound pretty darn good considering how much data is removed in the compression.

Bob L

[demodoc]

I generally don't go anywhere under 192kbps. Actually, I like 320 kbps...certainly the best quality and nobody ever complains about file size to me.

[mako]

I do all my backings at 192.
320 is better though if you want the next step.

cheers

mako

[Ian Alexander]

What are you using to make the mp3 files? Some programs have choices of encoders, speed vs. quality, etc.

[bcorkery]

I like 192 for stereo and 128 for mono files.

[alchemist]

Long ago some people made a big test of MP3 encoding bitrates with audiophile playback equipment and "golden-eared" people, aiming to discover at which settings the resulting MP3s were indistinguishable from the originals.

They ended up with the -r3mix settings on the LAME encoder. With that setting the resulting files were indistinguishable by the listeners from .WAV files.

But that was a long time ago, and the LAME encoder has improved a lot since that happened. The LAME encoder has now the --alt-preset-standard setting which is superior to the old -r3mix setting.

PLEASE Don't do constant bitrate encoding (CBR). It is a huge compromise. For some kinds of low average volume music (classical, jazz) it can be OK, but for rock it is awful.

128 kbps CBR - terrible, you can hear the artifacts even from far away
192 kbps CBR - somewhat less harsh, but you can hear the high-end artifacts on passages with high volume.
320 kbps CBR - excellent, but it is OVERKILL, low volume or silent passages don't benefit from 320 kbps and your MP3 files will be huge. It misses the point.

The solution? Don't encode with CBR. It is really easy to encode top-notch MP3 files.

Do the following:

1) Download the free WinLAME encoder at http://winlame.sourceforge.net/

2) Run WinLAME

3) Add the WAV files to be encoded to the queue

4) Select the --alt-preset-standard setting (I think it is now just called Standard)

Done!!!

Listen to the resulting MP3 files ... awesome, aren't they?

[Carl G.]

I just started doing alot of MP3's for people. I am losing alot in the MP3 vs. the uncompressed wave. Any thoughts, as to what is the best bit rate, compression type or mix ideas to compensate for the loss in quality, is appreciated.

I have heard some good MP3s, I am just coming up a little short....

Regards,
Rich
320 always *
Minimum artifacts.
Most people don't know what their missing - until your competitor plays them something better. :)


* (lots of reasons to compromise - you can make a 'sound' decision from there to fit your need)

[Studio de' Lara]

Thanks for the input. I use Samplitude for final mastering / CD burning / mp3 creation.
I will try the lame encoder and compare.
Regards,
Rich

[Naturally Digital]

I have to react to the above comment about CBR. Constant Bit Rate is the only way to go if you are a pro.
If you are trying to save space and get another 1,000 songs on your iPod, OK.
But, when you want to be sure a client will be able to play your work, CBR is a more forgiving universal way to encode.
Try encoding something in video with VBR, and you'll have a syncing nightmare.
That's my worthless $0.02.I must add that I've always felt the same way as Stashu on this CBR vs. VBR issue. I have no facts or tests to back it up though... only a gut feel. I "thought" CBR was better and VBR was mainly used to reduce file size. I've never tested the difference because I've always used CBR. I've been happy enough with the results. I use 320Kbps unless the client needs mp3's for web or ring tones etc. Some clients require 128K but I try to get them up to 192K if possible.

If you need to create very small (low bitrate) mp3's, especially for ringtones then try reducing the bandwidth by filtering the highs and lows. Lose the octave above 10K (or 5K if necessary) and below 60Hz (or 100Hz) before you encode your mp3. You may also get decent results by reducing the samplerate to 22kHz (which will eliminate information above 10kHz due to nyquist).

[alchemist]

You guys are nuts. 320kbps sounds pretty good, but the filesizes it produces are so huge that you completely miss the point of encoding to MP3. If you're that paranoid, just give the client a FLAC or the WAV file.

I dare you to compare a 320kbps CBR against a --alt-present-standard encoded file. You will not hear a difference, even with audiophile headphones and amplifier. And the resulting files are much better.

Do the comparisons and listen before making your decisions instead of believing a fairy tale.

[Dave Labrecque]

I use 192 bitrate and straight stereo format... my mp3's sound pretty darn good considering how much data is removed in the compression.

Bob L

Me, too.

Also -- there is a big difference between the 'high' quality setting and 'very high' for the LAME codec. That's the one I use -- via dBPoweramp.

Note: some MP3 converters (notably JMS' LACE) do not allow access to the 'very high' setting.

[Dave Labrecque]

Industry standard in radio is either wav, or mp3 at 192 or 320. 256 seems to be problematic with some systems.

Interesting. The daily weathercasts that we guys do for the Radio Forecast Network (serving over 200 stations) mandates 256 kbps, 44.1 KHz, stereo files for upload to the server.

A bit rate incompatibility seems unlikely to me.

[alchemist]

Radio stations use MP3s? wow.

Well, yeah, if you're using ancient hardware or firmware CBR makes sense for compatibility reasons. Now that I think about it my sister had a first-generation CD-based MP3 player that didn't work with VBR files.

You're not crazy. I was only thinking about internet distribution for software playing or modern MP3 players.

[Demodave]

I send all radio spots out at CBR 256 using LAME in dbPoweramp. I have never had a station or client report a problem in several years. And the sound quality vs filesize is a good balance.

My personal music collection on the laptop is all ripped from CD's at VBR -V 0 using LAME on the latest reference version (R12.4) of dbPoweramp. The sound quality is fantastic, while the average bitrate is around 240 kbps. On a side note, the MP3's that Amazon is selling are CBR 256.

For professional use though, I stick to CBR because you never know what the client on the other side is going to use to play the spot. CBR works in everything, VBR does not.

[Carlos Mills]

My real world experience with Radio Stations:LAME encoder, CBR, 256 kbps (they asked for it); JMS LAME encoder.
As a side note, I did a comparison some time ago between LAME and SF's mp3 encoder and heard a really big difference, favoring LAME.


I send all radio spots out at CBR 256 using LAME in dbPoweramp. I have never had a station or client report a problem in several years. And the sound quality vs filesize is a good balance.

My personal music collection on the laptop is all ripped from CD's at VBR -V 0 using LAME on the latest reference version (R12.4) of dbPoweramp. The sound quality is fantastic, while the average bitrate is around 240 kbps. On a side note, the MP3's that Amazon is selling are CBR 256.

For professional use though, I stick to CBR because you never know what the client on the other side is going to use to play the spot. CBR works in everything, VBR does not.

[Dave Labrecque]

My real world experience with Radio Stations:LAME encoder, CBR, 256 kbps (they asked for it); JMS LAME encoder.
As a side note, I did a comparison some time ago between LAME and SF's mp3 encoder and heard a really big difference, favoring LAME.

You should try LAME's very high quality setting. It sounds even better. :cool:

[Ian Alexander]

You should try LAME's very high quality setting. It sounds even better. :cool:
Probably even better than the original. I've also heard that if you type the filename in green, you get better definition in the mids, shimmering highs, and incredible depth on the stereo image.:rolleyes:

For years, I sent 160k CBR mp3 spots to radio stations. I actually had a few stations that used the sneaker net to get files from the email computer to the production room computer using, are you ready, a floppy. 160k kept the spots under 1.4 MB.

I recently bumped up to 192 stereo for spots. Never had any complaints about 160. I just figured that floppies were gone and email in boxes were getting bigger, so I'd go for a lower chance of cascading issues.

After reading some of these posts, I made about 10 different mp3's of a spot, using bitrates from 192 to 320, CBR, and VBR, even stereo and joint stereo, using LACE/LAME on the highest quality settings. Then I decoded them back to wav and lined em all up with a wav file mixdown in the SS MT. I used hot track solo to switch from file to file. Any differences I think I heard we so subtle that I would never argue about it until after a proper double blind ABX trial.

Then I reversed the phase on the uncompressed wav and mixed it with the other decoded files, one at a time. None of the decoded files zeroed out the original, but I had to turn up the monitors to hear it. The difference between the decoded files was also very very small, which confirms that, to me, they sounded very similar to each other.

I have occasionally heard my spots on the air. (I mostly listen to NPR.) Even in the 160k days, I NEVER heard a glitch, a compression artifact, even a hint of swishy highs on the air. At 192, I feel completely comfortable with the quality and the file size.

[Dave Labrecque]

Probably even better than the original. I've also heard that if you type the filename in green, you get better definition in the mids, shimmering highs, and incredible depth on the stereo image.:rolleyes:


Ian,

Not sure, but it sounds like you think I was joking. Or not serious. Or delusional.

The MP3s I make, even at 256 kbps, using LAME's high quality setting (in LACE, for example) can have a pretty obvious distorted sound on transients. This goes away when I use the very high quality setting on dBPowerAmp (not available in LACE, as mentioned earlier).

I'm a little baffled that this seems to be of no interest to anyone. :confused:

Everyone's comparing bitrates and CBR vs. VBR, but no talk of the quality setting, which can be any number of settings. Which means that different folks could be talking apples, while others are talking citrus.

[Demodave]

Another factor to consider...

More and more radio stations are going HD. That means that the file you send to the radio station could get reduced to an AAC bitrate as low as 32 kbps for the digital broadcast. Many stations also use some type of data compression for the commercials (and sometimes music) on the automation system. That's two compression cycles that your commercial will go through before it comes out of the radio. That's why it is imperative to send a higher-quality file if possible.

[DominicPerry]

32 kbps. I see why it's called HD.

Dominic

[Dave Labrecque]

For lager projects...

Ah, yes... the fine folks in the cereal malt beverage industry. How do you court those gigs, exactly? :p

[Dave Labrecque]

32 kbps. I see why it's called HD.

Dominic

Yeah, I only recently got a primer on how the whole HD radio thing works, technically, from a buddy who's a local station group chief. Not a move toward hi-fidelity. :(

[Ian Alexander]

Ian,

Not sure, but it sounds like you think I was joking. Or not serious. Or delusional.

The MP3s I make, even at 256 kbps, using LAME's high quality setting (in LACE, for example) can have a pretty obvious distorted sound on transients. This goes away when I use the very high quality setting on dBPowerAmp (not available in LACE, as mentioned earlier).

I'm a little baffled that this seems to be of no interest to anyone. :confused:

Everyone's comparing bitrates and CBR vs. VBR, but no talk of the quality setting, which can be any number of settings. Which means that different folks could be talking apples, while others are talking citrus.
Yup, I was fooled into thinking you were joking. Knowing now that you were either serious or delusional, I will find an app that allows the very high setting and do some listening. Thanks for persisting.:)

I'm a little baffled that in these days of incredible cpu horsepower, quality settings other than the very best are offered. So it takes a few seconds longer to encode? I guess if Encoder A takes longer than Encoder B, no one will notice or care that B sounds worse.

[Dave Labrecque]

Yup, I was fooled into thinking you were joking. Knowing now that you were either serious or delusional, I will find an app that allows the very high setting and do some listening. Thanks for persisting.:)

I'm a little baffled that in these days of incredible cpu horsepower, quality settings other than the very best are offered. So it takes a few seconds longer to encode? I guess if Encoder A takes longer than Encoder B, no one will notice or care that B sounds worse.

Actually, there's a significant difference in the processing time between high and very high. At least on my ancient P4 hyperthreader. Like almost a factor of ten. :eek: So, I do find it handy sometimes to use the quicker setting for non-critical and draft stuff. Especially when there's lots of it do be done. ;)

[Carl G.]

(to dave) I will find an app that allows the very high setting and do some listening. Thanks for persisting.:)
....
Ian, Let me know what you find out.

[bcorkery]

How do you print in green? I want that feature!