//                             \\
       |!    /    0````````````0   \    !|
      |!     [  ( |  Speexenc  | )  |    !|
       |!    \    0............0   /    !|
        \\                             //


So here's how to use speexenc, the lil' encoder-that-could.

For this to work, you should have speexenc and sox installed on your linux system.

First of all, your input sound file should be in either 8khz, 16khz, or 32khz:

p------------------------------------------------q
| bash$ sox foo.bar channels 1 rate 16k baz.wav  |
b------------------------------------------------d

Note that I've also converted it from a .bar file to a .wav file.  .wav is a good 
all-purpose format that we may as well use since every audio application will
accept it as an input file.

I've also converted it to mono.
If you're going out to speex, you also probably don't need stereo sound.  This is
just an assumption on my part, but seriously if it's fidelity you seek then speex
isn't the codec you're looking for.

Now let's feed this to speexenc and make it sound bad.

p-----------------------------------------------------------------------q
| bash$ speexenc --author "klaatu" --title "my_speex_file"              |
|  --abr 64000 --denoise --quality 2 --comp 6 -w baz.wav foo.speex.ogg  |
b-----------------------------------------------------------------------d

Ok, step by step:

speexenc = the name of the app :)

--author = meta tag containing my name

--title =  meta tag contaiing track title

--abr =    Average Bit Rate. this is optional.  Another option is VBR, and 
           depending on a lot of different factors, you choice will vary on
           what to use.  Experiment, knock yourself out.

--denoise = A great option to reduce snaps, crackles, pops created by encoding

--quality = 1-10 I think; 10 being really high quality.  I use speex on my
            oggcast to create super small files, so I go very low qual.  But you
            may not want this, so..again, experiment.

--comp    = complexity.  This has a lot to do with how more 'complex' sounds are 
            encoded.  If there is music (and there really shouldn't be in a speex
            file) then you'll need really high complexity, like 6 or above.  For
            my oggcast I use 3, because I cut music from my speex version.  But
            for Information Underground, I use a high complexity because we don't 
            mind the larger file size and we do include some basic music.

-w        = "wideband", which is hardcoding the fact that Hey, we're feeding you a 
            16khz file!  Other options are -n for 8khz and -u for 32khz.  This is 
	    not technically necessary because speexenc will kinda auto-detect it.
            But I use it anyway.

baz.wav   = input file

foo.speex.ogg = output file. notice that i'm stuffing the speex file into an ogg
                container.  I do this because servers don't know what a speex file
                is, half the time, while they more often can be trained to recognize
                an ogg.  I just find the ogg container makes the file a little more
                universal, which is helpful when using an uber-obscure codec.

... and that's about it!


Notes:

Speex doesn't like music, or really any sound that is longer than the typical human
speech-produced long vowel sound.  If you have complex sounds, then use a high --comp
setting.

Variable bit rate is nice.  I hard code --abr when I don't care about file size, and
use --vbr when I like to make sure I'm getting the best quality in the smallest file.