KEEPING FILE SIZE DOWN WITH MP3/OGG

Keeping File Size Down with MP3/OGG ~ Having Your Cake and Eating It Too
by Great Red Spirit
Version 1


1. Introduction
So you're making a game, an epic tale of heroes versus the forces of darkness. It wouldn't be an epic adventure without some great music so you go looking for some. Before long you have your soundtrack, except there's one small problem: They're all bloated MP3s and including these tracks would drastically increase the file size of your game! Now's where you need to work on your musical finds to reduce their file size so that they no longer bloat your game's file size while still using the same great music you found. You'll need to edit it to cut out redundant data, if desired you can loop it better, and encode it at a lower bitrate where you can drop the file size significantly while preserving the quality as best you can.

1.1 The Advantages ~ How Low Can I Go?
This depends on various factors:
- What bitrate was the file originally encoded at?
- At what bitrate does the music start to sound bad?
- What audio codecs can you use?
- How many times does the original file loop?
- Is there any lead-in or lead-outs?

Here's one comparison:

Original MP3 Edited File
File Size: 4.98MB (5099KB) 0.66MB (678KB)
Track Length: 4:23 1:55
Audio Codec: MP3 OGG
Bitrate: VBR~160kbps VBR 49kbps
# of Loops: 2 1


5MB versus 0.66MB is a drastic advantage between the original file and an edited version. The returns can be better as some MP3s can be encoded as high as 320kbps, but the same applies as you might not get the same low bitrates with an acceptable loss in quality. There's still an advantage to be had by some simple editing.

1.2 Alternative ~ What About MIDI?
Audio codecs such as OGG and MP3 offer high quality music but at the cost but at a high cost. The alternative music format, MIDI, can't match MP3/OGG's quality potential but they have an great file size advantage. The longest MIDI's file size blow away short MP3/OGG's file size; A 16 minute MIDI file is only 225KB while a 3:44 MP3 is 5.2MB (5283KB). The file size advantage of MIDI can't be matched, but its possible to shrink the disadvantage and the file size of MP3/OGG formats. MIDI however can't match the quality potential of MP3/OGG.


2. Getting Set Up ~ What Do I Need?
Besides your music, you need two essential programs:
1) Audio editing software
2) Audio encoding software

For this tutorial I'll be using the following programs:
1) Windows Sound Recorder for editing WAV files
2) dbPowerAmp for encoding WAV files into MP3/OGG
3) Winamp to convert music formats to WAV files

Some alternatives to Windows Sound Recorder would be programs like Audacity or Goldwave. These programs are outside the scope of this tutorial and you'll have to try them out on your own, but they do have a lot more features than Microsoft Sound Recorder.

For dbPowerAmp, you need to download additional codecs available at their site to add support for OGG encoding. OGG Codec link. For MP3 encoding the built-in LAME MP3 encoder has a license that allows for 30 free days of use. A free MP3 codec called MP3 Helix. MP3 Helix Codec link.

3. Editing Your Music ~ Cutting the Fat
The idea is simple enough. You remove anything that is redundant to the music. This is usually things like silence at the start or end of the track, lead-ins or lead-outs that when the music loops sounds awful, and when the music itself repeats like most OST tracks do. The first step is to get your music in an easy to edit format if it isn't already. For this tutorial WAV will be the used format, and most music doesn't come in that format so the first step is to convert the music.

This is the track that will be used for this tutorial if you want to follow along. *2

Once it is converted into WAV, the next step is to cut out loops, lead-ins, and lead-outs as you so desire. You should listen to the music a few times and have an idea of what should be cut, such as any extra loops in the music or a fade out lead-out. Once you're done editing your music, the next step is to reencode it into a smaller file size.

3.1 Converting to WAV ~ Because Microsoft Sound Recorder Doesn't Support Anything Else
So how does Winamp convert music to the WAV format? Winamp has an output plugin that outputs whatever it is playing into a WAV file. First, turn off repeat in Winamp so it won't keep trying to write the WAV file when its finished. Next hit Ctrl-P to open the Winamp preferences.


You should see something like this. Click on Output under Plugins.


This is a list of all available output plugins for Winamp. Select the Nullsoft Disk Writer and click configure.


Set the output directory to whereever would be best for you and click OK. This will take you back to the output plugins, click close.

Now play the music track through Winamp. The progress bar will speed across since it isn't actually playing the music, just converting it. You now have a WAV version of your music track! This method works with most audio formats including exotic ones such as SPC. NSF formats that don't end have to be manually stopped. WMA formats aren't supported due to licensing issues with Microsoft. You'll have to find another way to convert to WAV or edit the WMA file directly which is beyond the scope of this tutorial.

3.2 Where and How to Cut ~ Beware Sharp Edges
The first thing to remove is any loops. Generally there's only one or two loops and our demonstration track has two. You can estimate the location of where the music loops by taking the total length and cutting it in half, but to find the exact point you'll have to play it by ear. The lead-in and lead-out length affect this since the track is usually Lead-In/Loop1/Loop2/Seconds of Loop 3/Lead-out.

For our demo track which is 224 seconds long, the loop point would be around 112 seconds (1:52). Its easy to pick it off for this track too, and I'd estimate it the loop to be around our estimated target. Now what you want is to cut out one of the loops. After all its just the music repeating itself, right? What loop to cut depends on what you want: The first loop has the proper lead-in and setup, but the second loop sounds better when its being played continuously. I generally shoot for the second loop, so I'll cut out the first loop.

Open up your WAV file in Microsoft Sound Recorder and go where you thought the sound looped:

Give the exact loop point some space though, this is a inprecise cut for now. You want to get the loop point and some of the unwanted stuff around it. We'll do a proper cut later. Under Edit go to Delete Before Current Position. This will cut out most of the first loop and lead-in.

Next step is to cut out the start of the third loop and the lead-out. Its almost the same as before: Go to around where the loop ends and listen for when the music starts to loop again. Its a bit before 95 seconds (1:35) so under Edit go to Delete After Current Positoin when you're at that point in the track.


Save this file! A different name would be best in case of a mistake, but now we'll work on making it a fine cut loop since its pretty coarse right now.


Move the seek bar to the start of the track like so:

Push the right direction key on the keyboard when the seek bar is highlighted like it is above. This will move the seek bar .10 seconds forward. Move it once or twice and cut everything before the current position. This will cut a small amount of the pre-loop music we still have. Keep doing it until you're happy with how it starts. Remember there is no undo so if you mess up and cut too much, reload your saved file from before!

When you're done with the start of the track, go to the very end and cut off bits from the end.

To make sure you got a good fine cut that loops good, import the WAV file into your game and play it through the editor music player or play it like a regular music file. Listen specifically for the loop point and make sure it sounds good. If you can notice the loop casually, go back to your coarse cut save and try again. Otherwise, good job! You just did the hardest part of this entire process.

Notice: If you use RM2k(3), after you save your fine cut, go under Effects and select Increase Speed (100%). Save this file under a different name.


4. Encoding to MP3/OGG ~ FORMAT WAR!
So now you have your clean cut WAV file. The next step is to encode this file using an audio codec. The WAV format uses lossless compression and no compression which gives it a huge file size that is neither wanted or needed. The main choices are OGG and MP3. OGG is the better compression codec of the two but it doesn't have the same support as MP3. If you use RM2k(3), you want MP3 and for RMXP/VX, you want OGG.

The idea is to compress your file using various low bitrates and to see which bitrate is your sweetspot: A good match of file size and quality. The fine cut demo track will be encoded using several different bitrates under MP3 and OGG.

Here is the fine cut version of the demo track that I'll be encoding at low bitrates. Its a higher quality OGG so there shouldn't be any non-trivial quality loss from the original file.

To encode your file, open dbPowerAmp and select your fine cut WAV file. You should get a window similar to this:


The pull down menu at the top selects what codec you are using to encode at. The options discussed in this tutorial are MP3 Helix and Ogg Vorbis. There are other MP3 encoders available, such as LAME.

The target can be any option, but this tutorial used the selected one, variable bit rate (VBR) for all encoding.

The last highlighted part is the quality slider. Farther left means less quality and file size, while farther right means more quality but greater file size. This is what you want to play with to find the ideal bitrate to encode at that has both quality and file size. Pick a bitrate, convert, and listen to see if its quality is acceptable or not. Repeat until you find your sweet spot or you just can't encode low enough.

4.1 The Results ~ Winning Arguments on the Internet
Here's the file outputs of the fine cut input file. The MP3 was encoded using MP3 Helix and OGG was encoded using OGG Vorbis. The lowest MP3 variable bitrate was 80kbps. There was a 48kbps constant bitrate under MP3 Helix, but it just encoded it at 96kbps constant.

Note: Bitrate is what is specified in dbPowerAmp, though Winamp reports higher.

MP3 OGG
VBR~120kbps VBR~128kbps
VBR~100kbps VBR~98kbps
VBR~90kbps VBR~80kbps
VBR~80kbps VBR~64kbps
VBR~48kbps

The sweet spot is purely subjective to what you want and if you can find any differences between the higher and lower bitrates. OGG supports lower bitrates so smaller file sizes are possible with OGG just through supported kbps. Find your favorite quality/file size bitrate and use that to encode your music and import into your project.

The smallest file, OGG VBR 48kbps, is 553KB big. The original file is 5285KB, nearly ten times bigger than the smallest OGG file! The returns of properly editing and encoding MP3/OGGs can be drastic at little quality cost. The return rate will be different for each file, depending on how it was originally encoded, if there's more than one loop in it, and how low you can get the bitrate with it still sounding acceptable.

4.2 Cutting File Sizes in Half ~ For Desperate People
If you use RM2k(3), you can cut the size of a MP3 track in half. If you double the speed of the music file (See the notice in Section 3.2) and encode that file then during playback in RM2k3a half the speed that the track plays, you can get a decent playback of the original file. The quality takes a hit but it is a drastic reduction in file size of about half. If you're desperate to reduce the file size of your audio files this is a possible step to take.

5. Conclusion ~ tl;dr
Cut out the lead-in, lead-out, and additional loops from your music. The second loop, if there is one, loops better. Encode your music using MP3 or OGG at the lowest possible bitrate while still preserving quality. Differences in file sizescan be around a factor of 10.

6. Music Credits ~ I Didn't Make Any of This!

*1 A mass Final Fantasy remix compilation. Compilation composer unknown, original composer is Nobuo Uematsu.

*2 Demon Spear of Wild ARMs Alter Code F composed by Michiko Naruke


** I thought there'd be a bigger difference between MP3 and OGG. Last time I did this there was, but that was with LAME MP3 and a different track. Oh well.

author=WIP link=topic=1405.msg21756#msg21756 date=1214431732
Why is this a forum post? Please submit this to the main site.

I'd like to give it an actual look over first since most of my proofreading was mostly checking that the links worked. I'll probably submit it tomorrow or so.

Neok: Thanks, I'll add some alternatives to it.

Wait what? XP/VX support that? I haven't seen anything like that in the help file, is there a place with more info about that?

author=S. F. LaValle link=topic=1405.msg21973#msg21973 date=1214535116
There's a tutorial up on another website, RMXP.org maybe, can't remember. You use the WinVorbis utility to write custom tags onto your ogg file, titled LOOPSTART and LOOPLENGTH (in samples). I'll dig up the link

Found it: http://rmxp.org/forums/index.php?topic=46547.0

I'm going to fiddle around with this and see if I can get it to work. It'll be nice to get loops going with the lead-in intact.