Audio File Conversion
Editing Sound Files
Equalizing Volume Level
Flattening Dynamics
Batching Multiple
Audio Files
Note
on
Installation of Programs
Audio file conversion
Using the unix program sox , it is extremely easy to convert audio files from one format to another. Suppose you have the aif file, "song.aif," on your desktop and you wish to convert it to wav format. Open the terminal and type:
cd Desktop This changes your working directory to your Desktop. Now all you have to do to change the format of your files is type:
sox song.aif song.wav This will create a file entitled "song.wav" on your desktop. Note: it will do nothing to your original "song.aif" file. Sox never alters the original file; only the file it creates on the basis of the original.
Sox supports a number of different audio file formats, including not only aif and wav, but au, dat, gsm, and ogg, among others. (See the sox man pages for more information.) One format sox doesn't support is mp3. However, you can use another unix program, lame, to translate mp3 files into wav files. The following command will convert the file "Etude.mp3" to "Song.wav":
lame --decode Etude.mp3 Song.wav What about converting files to the mp3 format? Well, of course, you can easily do that in iTunes. But suppose you wish to do it in the terminal? Here we once again have recourse to lame:
If you desire to encode by bitrate (at, say, a bitrate of 192) try:
lame -h file.aif file.mp3
lame -b 192 file.aif file.mp3
For more options, consult the lame man pages (enter man lame in the terminal).
Editing sound files with Audacity
Editing sound files is very easy, regardless of whether you use Audacity or some other shareware or commercial program. Chunks of sound can be high-lighted, deleted, copied, and pasted as you see fit. Audacity is mostly useful for removing or adding, through cut and paste, blocks of sound.
Here's a screenshot of what Audacity looks like:
Audacity is an open source project and is still a little rough around the edges. But it is constantly being worked on and improved and is good enough for those with modest sound editing needs. For more ambitious projects, a shareware program such as Amadeus would probably be preferable.
Equalizing volume levels
Suppose you have downloaded a number of mp3 files off the Internet, all from different sources, and want to make a single CD out of them. It is likely that these files will have been recorded at different levels, so that some will be too loud and some too soft. You could open them up in an audio editor like Audacity and try to level them manually, but that would be a lot of work. A better method is to "normalize" the files with the command-line program normalize. This program will change the volume settings of each audio file so that they will all be around the same without affecting the dynamics of any of the files in question. Using the program is very simple. Open up the terminal and type normalize . Make sure to include a space after the word normalize. Then all you have to do is drag all the audio files you wish to normalize into the terminal and press enter. Just make sure the files are in .aiff or .wav format. MP3 files will not work with normalize.
You can also "batch" normalize — i.e., you can preserve the relative sound values of a group of sound files so that the dynamic range is preserved between them. This is useful if you are combining files from a single CD or album with other files. To quote the normalize "man" page:
In batch mode, all the specified files are considered to be part of a single album and their relative volumes are preserved. This is done by averaging the volumes of all the files, computing a single adjustment fromthat, and applying the same adjustment to all the files. Some analysis is also done so that files with volumes that appear to be statistical aberrations are not considered in the average. This is useful if you have albums (like many of the author's) in which there is one "quiet song" that throws off the average.
Using the batch mode is very easy. Simply use the "-b" option with every file you wish to be part of the batch. Then cease using it for files not part of the batch. For example, suppose you have five files you wish to normalize, file1.wav, file2.wav, file3.wav, file4.wav, file5.wav. Suppose further that the first three files are from the same CD. You commence your normalizing festivities with the following command:
normalize -b file1.wav file2.wav file3.wav
Then conclude with this command:
normalize file4.wav file5.wav
And you're finished! Normalize is the easiest of command-line programs.
Flattening dynamics
Sometimes it is convenient to flatten the dynamics of a soundfile. There may be a number of reasons to do this. Perhaps you desire to make a compilation disc for listening in your car. If there exists a wide dynamic range in the music on the disc, you will have trouble hearing the quiet parts over the noise of the engine. Or suppose you are making a home movie and wish add a soundtrack. Again, if you chose orchestral music from a soundtrack or a work of classical music, you may need to flatten out some of peaks and valleys in the music's dynamic landscape. Sox allows you to do just that with the compand filter, although the syntax is rather obscure. The soxexam man pages recommend the following settings to compand the classical music work Also Sprach Zarathustra by Richard Strauss:
sox asz.aif aszc.aif compand 0.3,1 -90,-90,-70,-70,-60,-20,0,0 -5,0,0.2
Here's what this terminal command does. It first takes the asz.aif file and outputs it as aszc.aif, subjecting it to the compand filter. The soxexam man pages (enter man soxexam in the terminal) explain the settings as follows:
The transfer function ("-90,...") says that very soft sounds between -90 and -70 decibels (-90 is about the limit of 16-bit encoding) will remain unchanged. That keeps the compander from boosting the volume on "silent"passages such as between movements. However, sounds in the range -60 decibels to 0 decibels (maximum volume) will be boosted so that the 60-dB dynamic range of the original music will be compressed 3-to-1 into a 20-dB range, which is wide enough to enjoy the music but narrow enough to get around the road noise. The -5 dB output gain is needed to avoid clipping (the number is inexact, and was derived byexperimentation). The 0 for the initial volume will work fine for a clip that starts with a bit of silence, and the delay of 0.2 has the effect of causing the compander to react a bit more quickly to sudden volume changes.If this account of the process is accurate, it would seem to suggest that the part of the function that is critical to determining the extent of dynamics compression is the -60,-20. In order to test this, I ran the following command :
Notice that I not only change -60,-20 to -50,-15, but I also changed the -70,-70 part of the function to -60,-60. The reason for this is that in quieter passages, the -70,-70 was creating a lot of gain noise. I also changed the output gain level from -5 to -1 in order to preserve the impact of the first climax. Visually, this is what the results were like. First, the original file, "fg.aif":sox fg.aif fg2.aif compand 0.3,1 -90,-90,-60,-60,-50,-15,0,0 -5,0,0.2
Notice the extremes in the dynamic range. The first compand filter, the one from the soxexam man page, produced the following output:
This has the effect of pretty much flattening out nearly all the dynamics — too much for music, but perhaps just right for voice. Changing -20 to -15 causes the 3-to-1 compression to be squeezed into a 15 dB range rather than a -20 dB range. To flatten the dynamics less, simply raise this number, say, to -30. Be sure,however, to keep the output gain at -5 to avoid clipping and static.
Now the results of the final compand filter:
This should provide at least a vague idea of how this filter works. It's probably a good idea to play around with a little bit until you understand some of its nuances and can get from it precisely what you want.
Batching Multiple Sound Files
Running a command-line tool like sox can be tedious if you have to do it one file at a time. Isn't the whole point of using the terminal is the power and efficiency that lies within it's unassuming surface? Certainly it is. But how can you tap into the batch power of sox when it contains no batch potential of its own, as, say, normalize does?
It's not too terribly difficult, if you only know how. The secret is to use the foreach command in the terminal. First, you must be certain you are using the tcsh shell. Terminals run different shells, each of which has it's own unique features. If you're on Jaguar or a previous OS X system, tcsh is the default shell; so unless you've changed it yourself (or someone changed it for you), you shouldn't have a problem. If you're on Panther, your default shell is bash. You'll have to switch over to tcsh to take advantage of this tutorial.
If you're confused about this, simply enter :
foreach—then you have a bash shell. Enter tcsh and you'll be on your way. To batch a directory of aif files, here's what you do. Enter the first line:
If you're output looks like this—
bash: foreach: command not found
foreach file ( *.aif )
The word file is a variable. You can use any word you like, even a single letter if you're too lazy to type. When you enter this phrase, you will get a "foreach?" prompt. Here we tell the shell what we wish to do with all the aif files in the terminal's working directory. Let us compand them as follows:
sox $file $file.aif compand 0.4,1 -90,-90,-70,-70,-60,-30,0,0 -5,0,0.2
Once this line is entered, all you have to do is enter end to set off the process in which every aif file in the terminal's working directory will be sent the sox command as configured above. One thing to keep in mind, however. When batching like this with foreach, sox won't work properly if your audio files have spaces in their names. Because it can be extremely tedious to get the spaces out of file names, I wrote this quick-and-dirty command-line program, spacer, which replaces all the spaces in names with a "_". The binary can be found here. Just install it in your /usr/local/bin directory.
Note on Installation
It has been brought to my attention that some of the readers of the site are not clear on how to install the programs discussed herein. I briefly discussed the installation on the homepage, and will info on it soon. All the Unix programs discussed on these webpages, unless otherwise noted, can be installed using the program fink, which can be found here. This includes sox, normalize, and lame. Audacity can be found here.
—Last updated February 19, 2004