Generates waveform images from audio file.
Requires pyffmpeg (beta version, available on GitHub), numpy and PIL.

Supports:
- Waveform view with left and right channels
- Waveform view with maximum and minimum level of all averaged channels
- Antialiasing (oversampling)
In rendered_examples directory you'll find rendered images using this script.

Will support:
- Spectral view

Won't support:
- Colors

Drawbacks:
- Needs to read whole audio file at least twice

TODO:
- Document the code (docstrings, comments)
- Implement spectral view
- Create nice command line utility

About rendered examples:
Track is "Unsquare Dance" by Dave Brubeck.
unsquare1* are stereo waveforms: above is left channel, below is right channel (absolute maximal levels).
unsquare2* are mono waveforms: above are maximum levels, and below minimum.
Files without sufix are normal waveforms with no filtering.
Files with "_aa" suffix are antialiased: It's almost the same as rendering waveform with width 20 times larger than intended, and downscaling it with bilinear filtering.
Files with "_avg" suffix are "averaged": it's roughly the same as antialiased but with applied contrast.
Files with "_hyb" suffix are "hybrid": a little averaged first, and then antialiased...