Controlling a Synth using a Neural Network in SuperCollider

This video demonstrates how to use a neural network to control a synthesizer that has 10 control parameters using just the 2 control parameters of an XY pad with the FluCoMa Toolkit.

0:17 demo
0:24 theory
3:45 begin coding
5:21 FluidDataSet
7:09 FluidBufToKr
8:38 adding data points to FluidDataSet
12:54 saving FluidDataSets to disk
16:41 training the neural network (FluidMLPRegressor)
21:04 saving the state of FluidMLPRegressor to disk
22:27 making predictions with FluidMLPRegressor
26:00 updating the MultiSliderView with the predicted values
28:31 next steps

32:27 triggering predictions on the server using FluidMLPRegressor's .kr method

To learn more about FluidMLPRegressor visit:
https://learn.flucoma.org/reference/m...
https://learn.flucoma.org/learn/mlp-p...
https://learn.flucoma.org/learn/mlp-t...

Starting Code: https://learn.flucoma.org/examples/re...
Complete Code (without MLPRegressor .kr): https://learn.flucoma.org/examples/re...
Complete Code (with MLPRegressor .kr): https://learn.flucoma.org/examples/re...

The Fluid Corpus Manipulation Toolbox (FluCoMa) is a software extension that enables programmatic sound bank mining with machine listening and machine learning within Max, SuperCollider, and Pure Data.

Website: https://www.flucoma.org/
Download: https://www.flucoma.org/download/
Discourse: https://discourse.flucoma.org/

Max: https://cycling74.com/
SuperCollider: https://supercollider.github.io/
Pure Data: https://puredata.info/