This code uses a set of instructions to build a markov chain in python and play it using FluidSynth Midi synthetiser. The instructions are generated from a MIDI file (.mid) and converted into json format using ToneJS Midi. Then, they are fed into a simple markov chain that generates midi phrases out of the ones from the original track, and playes them using pyfluidsynth.
In this example, Aphex Twin - Aisatsana is used to generate the instructions in the markov chain. You can download the track's Midi File from the MIDM database.
You need to have the fluidsynth framework installed in order to route MIDI in your computer. In order to use this pyfluidsynth library, you may install fluidsynth's version 1.1.x:
git clone https://github.com/FluidSynth/fluidsynth.git
cd fluidsynth
git checkout 1.1.x
mkdir build
cd build
cmake ..
sudo make install
Then, you are safe to install python bindings for it with pyfluidsynth. It is a good idea to test the different tests in nwhitehead's pyfluidsynth and check that there is no actual error. Other more commonly used packages such as numpy are needed.
The script is written in python3. You need to use an SF2 file, in this case an acoustic Piano Soundfont (included in the Piano_SF) for the samples. You might need to change the midi driver in the endlessplayer.py file, according to your platform (in TODO-list to make it more flexible). In case of MAC:
self.fs.start(driver= "coreaudio", midi_driver="coremidi")
To start the player:
$ python endlessplayer.py instructions/instructions.json soundfont/SC55PianoV2.sf2
Parsing file: instructions/instructions.json
Title Aphex Twin - aisatsana
BPM 102
Generated markov chain. Playing.
Using SoundFont File Piano_SF/SC55PianoV2.sf2
NB: some extra tweaks are done in the json output from the MIDI Converter. Some interesting Metadata is added, such as the duration of the phrases or phraseLength.
Alex Bainter, for inspiring with his work in this post and in generative.fm. His code is writen in JS and this repository is nothing more than a translation into Python using different libraries. All creative credits are his (except for Aphex Twin's composition, ofc).
omgimanerd's markov-music repository, inspiring some of the functions for the Markov Chain creation.
nwhitehead's pyfluidsynth implementation of python bindings for FluidSynth. The endlessplayer uses a sequencer from this library as well.