[A6] RE: Neuron (ot)

[Rhen, Kris]

>      Ok, I'm going to ask a question, let's try to be 
> objective here. I
> personally care more about the end result of the synth (i.e., 
> the sounds it
> can make) moreso than the means (process it uses to make the 
> sounds) that
> it uses. So, ignoring the supposedly original synthesis 
> method the Neuron
> uses, am I the only one that thinks that any decent sampler 
> can be used to
> synthesize such sounds? 

Sampler - no.  The neuron is closer to something that does physical
modeling, like the Z1/Prophecy - only that those models are 'fixed' -
defined by the people who created them, so that, for example, a 'plucked
string' model has certain parameters (attack noise, damping, string
position, dispersion, etc)  The difference with the Neuron is that its
models are dynamic in that their 'parameters' are decided upon by the neural
net model making engine.

This is why there's not too many innovative pieces parts around the
oscillators (resynators).  MOST of what can happen with a sound is modeled
in the oscs - the filter is an added touch, BUT if you create your model
right, the filter properties become PART of the oscillator model.

Now I haven't received my modelmaker software (which was just released for
mac9 but not win yet and I don't have access to a mac) but that's kinda the
missing piece now.  In theory, you could feed it samples of a dog barking
and a duck quacking, and it would build a physical model of what a duckdog
creature would sound like allowing you to real-time tweak the model among
all the 'parameters' IT determines are happening within the source sample
set.  No sampler is going to do that.  Its beyond crossfades.  Hard to
explain until you really get your hands on one and set to the INIT PRESET,
pick a model, then start playing with the Sphere and Scape parameters - you
can tweak JUST one osc (no filter or efx) for hours and make a slew of great
sounds...

The nice thing about AI and neural nets is that they often come up with
'solutions' which are not expected even by the net programmer.  This leads
to unexpected cool things happening in your models.

To me, what's important in a synth is the VARIETY I can get out of it, as
much as the sounds end-result.  Especially since I'm an evolving paddy kinda
guy, the Neuron fits that bill quite well.  Hope I explained whats going on
inside the Neuron a little better with this, and once I get some quality
time making samples and making explanatory samples showingthings like just
what goes on in a raw resynator, that'll make things more apparent.  Maybe
this weekend, time permitting.
KRIS