supersational/JAMMIN-GPT

Ableton MIDI-Clip generation using GPT-4

JAMMIN-GPT

Paper | ISMIR Poster

Requirements

• abc2midi - https://abcmidi.sourceforge.io/
  • brew install abc2midi
  • if not installed using homebrew set path here: https://github.com/supersational/JAMMIN-GPT/blob/main/generate_midi.py#L15
• pip install openai>=1.1.1 rtmidi mido==1.3.0
• place 'client' folder in Ableton's Remote Scripts folder (for me it's in /Users/sven/Music/Ableton/User Library/Remote Scripts/AbletonOSC/client/)

Optional Requirements:

• fluidsynth (to use GM soundfonts)
  • run fluidsynth and select it as an output MIDI device in Ableton with 1 channel per track
  • fluidsynth path_to_soundfont/FluidR3_GM.sf2 -pFluidSynth is the command I use

Screenshot

Add AbletonOSC as a control surface in Ableton