MIDI Data Miner uses a neural network to learn correlations between notes and control changes in a MIDI file.
After training MDM can augment a live MIDI stream, adding control changes based on notes received.
Briefly: use MDM by connecting a MIDI device, open preferences and set MIDI input and output ports. Open a MIDI file, create a new neural net and train. Now play notes on your MIDI device, MDM will automatically add control changes.
MDM creates a collection of note and controller pairs from the file and uses them to train a net. The notes are used as inputs and the controls as outputs. The window is the number of previous notes considered when determining the output for the current note. Larger windows give less predictable more interesting responses. Enter a number in the window box and press enter to set the window size.
This is an early alpha release, buggy and incomplete. Please send feedback! Note that this alpha version only reads from the first track in a MIDI file, and only uses the first controller. Look at the included MIDI file in the working/midi folder for details.
· Java 1.5
What's New in This Release:
· The bug that stopped MDM from launching on most systems has been fixed.
· A window control has been added to allow the neural net to consider a window of previous notes.