pedal-machine
machine learning guitar pedal
summary
Pedal-Machine is a machine learning guitar pedal patch for Max 7. The pedal classifies which pedal effect to apply based on your current playing style. This project started out as a school project at UC Berkeley for my music technology course and it has been improved on since then. If you want to run the patch for yourself, make sure to use Max 7 and add the appropriate externals (Odot, CNMAT externals, OSC).
how it works
To achieve this, the patch captures two features of guitar playing: tempo and amplitude. It sends the current value for each feature through an AdaBoost classifier which then fades to the deduced pedal effect. It is straightforward to add more features if desired (like average frequency or detecting chords vs. single notes) but new training data must be included. In my demo, I walk through the patch and show it in action. During the video, I improvise on guitar while the pedal matches my playing style with a relevant pedal effect. When I play loud and fast, it adds an overdrive effect, and when I play slow and quite it adds a reverb effect.