Pratice in AI × music in python
(get into your virtualenv)
cd pyalgomuse
python3 setup.py develop
pip3 install -r requirements.txt
- TheoryTab Leadsheet Dataset
Thanks for one of the MuseGan's authors,he has done a lot of parsing work on TheoryTab dataset.I fork and modify his source code into pyalgomuse/utils/theorytab module
- piano-rs
A rust based program for playing virtual piano by computer keyboards in terminal.I use it for melody input.
[UNDER CONSTRUCTION]
Reasons:
- There is little Chinese/Asian songs in current public datasets.
- Lack of systemic tags in current datasets.
- No compatibility nor unity of constructing/parsing processing among different datasets
To generate corresponding chords for the given melody
- training: (The folder has already had trained results included,this step can be skipped)
put the <hooktheory_dataset> folder into a folder named datasets in root directory
cd scripts
python3 train.py
(the trained results store in A.npy,B.npy,symbol2idx.json)
- fitting:
cd scripts
(set your piano-rs path to $PRS Variable in record.sh)
./run_aa.sh
--- running steps ---
1.input your params of metronome,e.g. bpm & time_signature
2.play your melody on the active terminal with piano-rs,pay attention to follow the metronome.Press <Esc> to finish.
3.press <q> to quit the metronome.
4.press <y> to continue to listen the midi combination of the generated chords (column chord for simple playing)
and your melody input(1 octave higher than your actual input)
[UNDER CONSTRUCTION]
[UNDER CONSTRUCTION]
- follow the work of MuseGan
[UNDER CONSTRUCTION]
To generate melody by images,texts,videos ..or random input
[UNDER CONSTRUCTION]
To transfer a song into pop/classical/jazz/... style
[UNDER CONSTRUCTION]