Looking to read in massive amount of MIDI files and extract their data into groups - tick(time), pitch(note) and velocity(volume). The extracted data will be applied to a Recurrent Neural Network for further music generation. I have been searching for libraries and located python-midi. Here's my code: import midi pattern = midi.read_midifile("pachelbel_canon.mid") #print out all events print pattern #print out all events at tick = 0 print pattern[0] Interpretation of different event types can be found either on here or Granroth-Wilding's website. My target values (tick, pitch and velocity) are presented as midi.NOteOn/OffEvent(tick=0, channel=0, data=[pitch, velocity]). Question 1: Since there's no tutorial nor documentation available, after some testing I figured print pattern[0] returns all events happened at tick == 0. Of course all data can be retrieved by running a for-loop over all events, I'm just wondering is there any function to retrieve data by providing other parameters e.g. channel, pitch etc. Question 2: Apparently there are lots of midi / music libraries available out there. Not saying python-midi is not decent, yet if you have anything better in mind please do share! Preferably comes with some proper documentation or support, which doesn't exist for python-midi. I'm new to this area and would love some recommendations. Continue reading...
