Audio2png

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...