Python 3
glob for python (included in most distributions)
ffmpeg (binary for mac included)
pydub for python (pip install pydub)
Added Wavenet for audio waveform generation using tensorflow. So... it turns out that music is not as patternable as I previously thought. Turns out that there are too many caveats in music (even in the EDM corpus I used) to accurately generate audio. The generated audio is presented as a waveform and is basically just static--in other words, there's not discernable pattern so far. Perhaps, I need to use one type of genre or such.
Added corpus_gen.py and ffmpeg. ffmpeg is a unix binary built for Mac OS X to convert audio compression formats. corpus_gen.py is the initial dataset creation script. Basically, place all mp3 files scraped from youtube and other forms (if using youtube, I recommend using a playlist and downlaoding en masse using downvids.net - this can be sped up usig a simple jquery select to open every generated download link at once, and then the output mp3 can be cleaned using a simple bash loop to rename song titles so that it appears in pretty form without spaces) into the MP3 folder. I have not included mp3 files here due to copyright claims. Then, run the script and the corpus folder will have generated waveforms.