?M Fra*

The kernel of ?M Fra* is a set of files that were generated by Marco Pasini’s musika software from a dataset of audio files by Ian M Fraser and myself. Like teratomas — rare tumors that can sprout fully formed teeth, hair, bone, or other somatic structures — the output of this system featured familiar musical figures jutting haphazardly out of a more primitive, inchoate sonic mass. They register less as hybridizations of Ian’s work and mine than as para-musical found objects: stylistic markers from the corpus of royalty-free techno music used to train musika are audible, distended and compressed by the opaque compositional logic of the generative adversarial network, while the timbral dimension was characterized by a fuzzy indefiniteness that gave the impression of distance, that the sound was partially inaccessible due to the interposition of some occluding medium, never fully present but merely overheard in its state of virtual elsewhereness. These files were subjected to various source separation tools, each of which imposed its own spectral gestalt. Although the term source separation implies a genealogical protocol, a retrograde movement toward earlier, antecedent elements, the encounter between the musika products — a sonic object generated holistically rather than by summing discrete instrumental tracks — and the source separation algorithms yields novel entities, false histories ontologically posterior to that of which they are the putative “source.” By recombining stems from different source separation methods, the pattern recognition function of the various machine listening paradigms is subverted: new patterns emerge as a result of the juxtaposition of non-complementary stems, while reciprocal regions are riddled with gaps and spectral remainders. – RMF
https://www.librarystack.org/m-fra/?ref=unknown