Giving you musical superpowers
Muzoti is the fruit of a 27 year personal obsession into what makes music sound, well... musical.
A music score can, to some degree, be reduced to a sequence of numbers, so why is it that one sequence sounds good, yet another sequence of randomly generated notes sounds terrible? What inherent structure in a musical pattern that makes it sound good.
This is the question that I longed to answer. As the years went by I constantly tested my theories. I wrote programs that generated 'music' according the organising principles that I observed. The more musical the result, the more sure I became that I was on the right track...
Eventually it came down to this: if our brain has evolved to predict patterns in space and time, and rewards us when we get it right, what would happen if you turned it on it's head? What would happen if you tried to maximally stimulate the brain's reward system by showing it generated patterns that alternate between repeating and varying material, not just on one, but on many temporal scales? If those patterns were patterns of notes, would that sound like music?
Muzoti is an exploration of that thesis, and poses a fundamental question. Is music, or at least a substantial aspect of it, simply a by product of the way the brain has evolved, rather than a purely cultural phenomenon?
Unlike other generative music attempts, we're not using machine learning. Instead we are developing a solid basis for generative musical intelligence based on what we believe are universal organising principles of music and life. Our algorithms are style agnostic. Without such a solid framework, machine learning can only extract style information based on pre-comosed works. We believe machine learning, while valuable, should come second to a holistic theory of what music is and why it exists.