My next art project…sane?

Given an original piece of music, slice it into arbitrarily small time slices and perform FFT. From a corpus of (1000+) copyrighted songs similarly sliced, find the slice whose frequency content most closely matches the original slice. By pulling in such slices in place of the originals, reconstruct the piece, still sounding recognizably like the new piece but actually containing NO ORIGINAL MATERIAL. Bonus points: calculate how many millenia you could go to jail for for downloading this piece.

Variations: Allow each slice to be replaced by (smallnum >1) slices added together, after applying bandpass filter to isolate only the desired frequency component(s). Amplitude shaping and possibly frequency scaling may also be called for.

Perform the same project on images instead of music.

One Response to “My next art project…sane?”

  1. Steve Witham says:

    Mostly done, although with a different slant: .

    His slant is to start with a smaller source including video (e.g. one rock video or tv interview), take input from a mic, and in “real time” (1-2 seconds?)
    construct something *with video* that sounds kinda like the mic input. But his technique (closest-match of sound signatures in a database of fragments) sounds adaptable.

    Definitely watch the intro/ demo video first:


Leave a Reply