Skip to content
Use ML to compose music (midi format, note, chords, rests, durations)
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Generated
vocab_save
weight_save
.gitignore
LSTM_Midi_Generator_multitrack.ipynb
Magic_Piano_from_file.py
Magic_Piano_from_midi_keyboard.py
README.md
requirements.txt

README.md

Midi_Deep_Composer

Use ML to compose music (polyphonic, multitrack, midi input)

Generated examples:

https://soundcloud.com/klhnikov/sets/ai-lstm-generated-piano-parts

Magic Piano : play 32 notes, the piano answers...

python Magic_Piano_from_midi_keyboard.py

or 

python Magic_Piano_from_file.py [MIDIFILE]

Warning, you can only put 32 notes in the file... To do so, you can use this service, fill every note and save your song to a midi file : https://musiclab.chromeexperiments.com/Song-Maker

Data:

For each track, for each note, chord or rest, we used the pitch, duration, offset and velocity as "one-hot encoded" values.

We generate a note/chord or rest based on the seq_len (32) last notes/chords or rests. Long chunks of music are generated by sliding the window over time.

[A3, C5, [C2, E2, A2], rest, ...] == (seq_len notes) -> prediction -> predicted note

You can’t perform that action at this time.