Plugin thoughts
So, lately the DAW plugin market has been a bit saturated with a bunch of vendors all producing plugins that more or less do the same thing. This is great insofar as it’s driving down the price of plugins, but not so great insofar as it suggests no one can think of anything new to do.
My first and most obvious suggestion on this front is to abandon the beaten path of what’s been done before. In particular, might I suggest that there’s a huge GPU on most modern computers that could be used to play with building artificial neural networks to perform audio processing. I’m not exactly sure what the results of this would be, and maybe a GPU isn’t big enough yet to do meaningful amounts of ANN processing, but I still think it’s a neat idea.
I have memories when I was a tween of having a vision of manipulating a giant artificial neural network via a set of controls very similar to a mixing board, which controlled trigger levels and amplitudes of signals passing between subnets. I didn’t fully understand what I was imagining at the time, but looking back…
July 28th, 2016 at 10:32 am
That would be kind of fun to build… say, with Parallax Propeller MCUs, each programmed with 64(+?) neurons with an 8×8 grid of ‘weight’ controls and accompanying input sockets. Each core runs 8 neurons, perhaps with 8 outputs per square (currently unsure of per-core neural “capacity”). Of course, a method to take the resulting net and put it into a system without those external weights would be ideal. Would be a VERY interesting way to interact with and understand the results from an ANN.
Would also be expensive, not for the MCUs, but the POTs and sockets.
But now I really want to build at least one 8×8 square…