Method to make a MP3 small and still sound good?

[santo]

I am mixing a compilation of my songs in Pro Tools to playback looped (about 3min) on a Flash website. I am starting with a 48K 24bit mixes (premastered).

I am trying to get the MP3 file size down but still maintain reasonable audio quality.
On the master buss, I am compressing the compilation with a Waves SSL compressor, then into Ozone to compress even further, and finally into a Waves L1.

The file size is still a little too big. Wondered if anyone had some ideas or experience getting good results.

Thanks in advance.
Santo

[Mixocalypse]

Originally Posted by santo

I am trying to get the MP3 file size down but still maintain reasonable audio quality.
On the master buss, I am compressing the compilation with a Waves SSL compressor, then into Ozone to compress even further, and finally into a Waves L1.

The file size is still a little too big. Wondered if anyone had some ideas or experience getting good results.

The kind of compression you're talking about does not reduce files size? OR Im i totally misunderstanding you? Its sound like you thing that adding more mixbuss compression will reduce file size? ?

For Mp3 conversion for flash w/ good quality you wouldn't go any lower than 128kbs. 160 - 192 would be better.

[santo]

Yes I thought additional compression would reduce the file size. So if that does not, is there anything that will?

Santo

[Mixocalypse]

A different mp3 compression size. ie.. 80kbs 96kbs, 112,kbs, 128kbs, 160kbs, 192kbs
example.. 80 or 96 is what myspace plays its mp3s back at. blahhh.
the smaller the conversion settings the smaller the size but at the cost of sound quality.

Maybe some one else can chime in with something new..

[luctellier]

Use variable bitrate MP3 compression.

[tonerowlabs]

first of all, i think you are confusing DSP compression, which manipulates audio, with file compression, which manipulates file size using bit depth and sample rate. Totally different. So the first thing I'd do is choose ONE of those mastering limiters and get rid of the other two - that's not gonna help and it's gonna sound like crap.

In fact, the "thicker" the mix, the worse it will sound compressed and streaming in Flash. Overcompressing the audio is a terrible idea, i think.

Mp3 sampling rates can be chosen in the Bounce menu. I'd recommend using a Dither plugin (protools comes with an okay one) which will make it sound better at lower rates.

[santo]

Use variable bitrate MP3 compression.

I've noticed the Pro Tools MP3 export does not have a varible rate option, is that true. Or, am I missing something?

Santo

[Stisse]

Originally Posted by StaticStudios

A different mp3 compression size. ie.. 80kbs 96kbs, 112,kbs, 128kbs, 160kbs, 192kbs
example.. 80 or 96 is what myspace plays its mp3s back at. blahhh.
the smaller the conversion settings the smaller the size but at the cost of sound quality.

Maybe some one else can chime in with something new..

Actually I was really amazed when I started listening to Swedish Radio (www.sr.se) and their streamed radio. They broadcast with 96k but I could've sworn it was a really good 128kbit. If you have a chance, go there and have a listen.... Can't really understand why so many online makes a mess when apparently it's so easy to make it right....

Take care,
Stefan

[Will]

Different codecs sound different too, so I'd get a few different codecs together and make comparisions between them (so encode a track with the pro tools codec, then one with quicktime with the same settings and see which one you think sounds better).

[treymonfauntre]

export everything as 16/44 wavs, and use dbpoweramp converter to convert them to whatever you want. if its a flash player you probably don't want to go higher than 128vbr.. most people don't have their soundcards hooked up to good speakers.

[luctellier]

I use WaveLab to create my MP3s, lot of options, good compression, good quality. Never really tried another one.

[santo]

Thanks all. Investigating Options.

Santo

[tonerowlabs]

What do people recommend for Mac Intel? Anything noticeably better-sounding than iTunes (say, 128kbps, VBR at "Highest Quality")?

[Stevil]

if you really want to cut file size, make it mono.
it'll chop yer file size in half & you can afford to retain some bitrate. if you dont have alot of stereo/panning action in your mixes most people accustomed to internet quality music wont mind the difference.

[santo]

if you really want to cut file size, make it mono.
it'll chop yer file size in half & you can afford to retain some bitrate. if you dont have alot of stereo/panning action in your mixes most people accustomed to internet quality music wont mind the difference.

That's interesting, hadn't thought of it.

Santo

[theblue1]

My walters-detection circuitry is lighting up, here.

[theblue1]

Originally Posted by Stevil

if you really want to cut file size, make it mono.
it'll chop yer file size in half & you can afford to retain some bitrate. if you dont have alot of stereo/panning action in your mixes most people accustomed to internet quality music wont mind the difference.

Not strictly true... while you can use an algorithm that forces discrete stereo channels, many popular formats attempt to further reduce data size (or increase quality per bandwidth, if you like to look at it sunny side) by treating stereo and mono content differently, even switching algorithms as the information changes.

Here's some decent info on some of the issues. The author is a podcaster ( ) but he seems to know a lot more than a lot of recording "engineers" about perceptual encoding formats. (Actually, that really shouldn't be surprising. A lot of music recordists consider lossy, perceptual encoding algos to be the devil's handiwork -- but they are precisely what made podcasting possible.)

[treymonfauntre]

wheres the decent info?

[theblue1]

Originally Posted by treymonfauntre

wheres the decent info?

I thought -- for an instant -- you were calling me to task for that characterization...

and THEN I realized I hadn't actually put the damn link in.



http://www.blogarithms.com/index.php/mp3secrets/



If you spot any errors in his write-up, let me know. It all looked pretty solid from my quick scan.

[theblotted]

i do VO/SFX editing, mixing, and easy mastering for an internet flash animation website.

i give the producer 160kbps in mp3 format, which is the lowest i'm willing to go for best quality/size ratio. i use LAME encoder in Peak 5.2, but i also used Barbabatch with great results.

i wouldn't suggest using VBR, because i've found that it messes with phase at times. at worst it'll make some reverb tail sound like it's underwater. if it's a simple podcast where's it's mostly talking, and size is a bigger issue, do it. otherwise for music, i wouldn't touch it with a 10' pole.

i prefer AAC than mp3 anyday.

hope that helps some.

[santo]

i use LAME encoder in Peak 5.2, but i also used Barbabatch with great results.

Thanks for the sugestions. I will check it out.

Santo

[santo]

i prefer AAC than mp3 anyday

please excuse my ignorance. how do you encode with AAC? And will the flash developer be able to implement the file.

Do I buy Quick time Pro 7 from Apple?

Santo

[Mixocalypse]

Originally Posted by santo

please excuse my ignorance. how do you encode with AAC? And will the flash developer be able to implement the file.

No, use mp3 for flash development.

aside... AAC sounds better at smaller bit rates. ie.. a 160kbs aac will sound like if not better than a 192kbs mp3. and is a just much better sound codec than mp3.

But mp3 is more widely used by ALOT of players

[Jens]

Originally Posted by tonerowlabs

What do people recommend for Mac Intel? Anything noticeably better-sounding than iTunes (say, 128kbps, VBR at "Highest Quality")?

Soundconverter!

http://dekorte.com/projects/shareware/SoundConverter/

Probably some of my best spent 10 bucks ever.

Uses LAME and besides MP3 you can convert almost anything to anything. Great!

/Jens

[Stevil]

from the 'blogarithms.com/mp3secrets' blog

Assuming you’re encoding in Constant Bit Rate (CBR) format (recommended), the bit rate will determine the size of your output file relative to the length in time of the source. That ratio will be fixed and independent of whether you encode in mono or stereo. Stereo never generates a larger MP3. If anything, it reduces the quality instead.

this i would consider not strictly true.
my mono files are about half the size of my stereo files. it is half the data.
stereo is 2 channels so it's one set of data for left & another for right.
mono is one set so it's half the data. is he encoding a mono signal as a stereo file?

his argument as i read it is that by doubling file size to stereo:

I’m giving the encoder some help in doing its job. Okay, maybe I’m compulsive, but I believe I can hear the difference.

.

the goal here is to reduce file size, not to give encoders a helping hand.
yes you are losing quality but that is what the entire mp3 technology is based on, sacrificing signal to make smaller file size.
choosing what bits, signal paths & frequencies you are willing to sacrifice in the name of accessability & sound quality.
the average internet listener wont hear the difference between a mono or stereo signal on most tracks.
most club PA systems are setup mono.

Originally Posted by StaticStudios

No, use mp3 for flash development.

aside... AAC sounds better at smaller bit rates. ie.. a 160kbs aac will sound like if not better than a 192kbs mp3. and is a just much better sound codec than mp3.

But mp3 is more widely used by ALOT of players

That isn't necessarily true. While the container format MPEG-4 has potential, AAC encoders, especially from Apple, are still behind the development of MP3 encoders.


If you want to use mp3, you will need to set up LAME as am external encoder. Then you will want to use VBR (setting vbr-new) to get the best results. The reason is that the encoder can adapt to parts where more bits are required. CBR shouldn't be used anymore (Disclaimer: Unless you are shooting for most compatibility)
Another setting that you need is "joint-stereo".
While regular stereo forces the stereo, "burning" bits, "joint stereo" uses different stereophonic techniques, investing bits somewhere else, where it's needed more.

On the general quality issue:
The goal of lossy codecs is NOT to preserve the data, but to preserve the transparency. What does that mean?
Transparency for you is reached, when you can not (blind-test ABX) distinguish between the orginial WAVE and the MP3 file.
For most people, a VBR quality 2 mp3 using the latest LAME codec will be transparent . You might need a better quality setting, if you know how the encoder works and you listen for artifacts with an concious efford.

A very popular myth is, that mp3 "sound" different then the orginal. If properly encoded, this is not true. "Tinny", "Boomy", "Bass lacking" or other sound effects are psychoacoustic, because the encoder doesn't work that way.
What enables you to hear it is a mp3 are the so called artifacts, usually a highband ringing and such (samples are around somewhere). On which compression levels these become obvious, is individual.
This is the only difference. Stereo imagine and frequency response of the song is not altered by the encoder. At least by modern, VBR LAME.


While MP3 is fine, if you go an average of 192kbp/s an more, you should use MPC (Musepack) or OGG. Those codecs are more modern and better suited for high bitrates. Both of those use modern VBR mechanisms.

AAC and WMA, the commercial codecs, excell at bitrates around 128kbp/s and less. When going for transparency, high bitrates, it is currently not recommended to use AAC or WMA, as MPC and OGG are vastly superior.


For archiving purposes of CDs, a losless codec is recommended. It keeps the gaps alive and you can transencode easily in other codecs and also burn the CD in original quality. FLAC and APE are such formats.



Bottomline: Even most pros do not know enough about audio compression and approach it with a lot of mysticism.
Knowledge of how the encoders work enables you to choose the right quality setting and format.
If you have further questions, go ahead and ask.

[Adebar]

Originally Posted by zwaps

While MP3 is fine, if you go an average of 192kbp/s an more, you should use MPC (Musepack) or OGG. Those codecs are more modern and better suited for high bitrates. Both of those use modern VBR mechanisms.

AAC and WMA, the commercial codecs, excell at bitrates around 128kbp/s and less. When going for transparency, high bitrates, it is currently not recommended to use AAC or WMA, as MPC and OGG are vastly superior.

That´s what i think too from own experience and studiing the technologies.
I wonder why MP3 is still around. This codec is antiquated since years.

AAC is newer but also not young anymore. It came out - i think - about 1998. It is really far better than MP3 at lower bitrates. Everybody can ahve acces to it via iTunes, every iPOD uses AAC.

[santo]

Thanks for the reply, I do have more questions.

My bottom line is the success of this project is depended on a 3.5 minute music presentation. Encoded straight out of pro tools at 128, the file size is 3.3 MB. At 160 it is about 4.1 MB

Both take longer than I want to load during the flash presentation.

I think I need to have a 3.5 minute file at about 2.5 MB to load fast enough to keep people on the site.

What is the best way to achieve that and maintain resonable quality. (Step by Step if you have the time to tell me)

I think now I will tell the developer to give the website viewer the option to listen at high and reasonable quality (faster load)

So, given that. How do I encode for the very best quaility. And one at reasonable quality with a fast load?

I am starting with a 48KHZ 24 bit master mix.

Thanks,
Santo

I really don't like to dismiss people on forums, but other have said it so much better, I suggest you start reading here:
http://wiki.hydrogenaudio.org/index.php?title=LAME
And continue until your head explodes :D

Originally Posted by Adebar

That´s what i think too from own experience and studiing the technologies.
I wonder why MP3 is still around. This codec is antiquated since years.

AAC is newer but also not young anymore. It came out - i think - about 1998. It is really far better than MP3 at lower bitrates. Everybody can ahve acces to it via iTunes, every iPOD uses AAC.

Well, the codec itself itself is still in development with LAME. The container is not. But AAC is currently not better than mp3, except for low bitrates.
Apple and company have a lot of work todo.

[theblue1]

Originally Posted by Stevil

from the 'blogarithms.com/mp3secrets' blog


this i would consider not strictly true.
my mono files are about half the size of my stereo files. it is half the data.
stereo is 2 channels so it's one set of data for left & another for right.
mono is one set so it's half the data. is he encoding a mono signal as a stereo file?

his argument as i read it is that by doubling file size to stereo:
.

the goal here is to reduce file size, not to give encoders a helping hand.
yes you are losing quality but that is what the entire mp3 technology is based on, sacrificing signal to make smaller file size.
choosing what bits, signal paths & frequencies you are willing to sacrifice in the name of accessability & sound quality.
the average internet listener wont hear the difference between a mono or stereo signal on most tracks.
most club PA systems are setup mono.

I had to read that twice, myself -- BUT what he's saying is that a 128 kbps file is a 128 kbps file -- whether or not it's mono, stereo, or one of the "adaptive" formats.

kbps (kilobit per second) is a measurement of BANDWIDTH REQUIRED TO TRANSMIT -- NOT a measurement of quality.



Now -- at a "given quality rate" -- yes, a discrete stereo format will be roughly TWICE the size...

So if you're comparing a 128 kbps MONO file against a discrete stereo file (two discrete channels -- ie, "full" stereo) -- to get the SAME quality, you'll need to make that stereo file a 256 kbps file (assuming all else is the same) in order to get the same fidelity/quality.

See?

___________


PS... I'll second (or third or fourth or fifth, I guess) the recommendation of the LAME encoder. I think it produces excellent results -- particularly in high VBR settings. (This is NOT to say that the diff is night and day between MP3s encoded with LAME and Mp3s encoded by the "official" Fraunhofer codec (Fraunhofer developed the original MP3 codec and charges money for THEIR product -- which is why you often see a "free trial" encoder included with many programs -- but to continue using the Fraunhofer codec, you'll have to shell out something like $20-$30 in most cases -- even though the vendor/software publishers are only charged $3 a copy (if my info is up to date)... Believe it or not, the extra money supposedly just covers the cost of dealing with Fraunhofer to get the license. Having bought a $50 codec package from Fraunhofer/Thomson I can assure you that they are UTTERLY CLUELESS as business people (or were 6 or 7 years ago) so I wouldn't be surprised.

Last time I sat down and did a head-to-head between Fraunhofer and LAME, I became convinced that -- at bitrates of 192 kbps and up, LAME had the edge. Maybe barely but I preferred the LAME encodings.

That said, LAME IS SLOOOOW... which is probably not that big a deal on today's computers but 5 years ago I often found encoding a 3 minute song at the slowest/best quality of processing could take 15 minutes or more.

Many DAWs and other audio apps will allow you to "plug in" different codecs. (Sonar, for instance, has a codec manager utility that will allow you to set up command line parameters for the LAME codec that will allow you to use it as a "pushbutton" app out of the DAW.)

While the LAME codec has a DOS-like command line codec, you can get a lot of "wrapper" applications (like the old CDEX) that will put a GUI on it.