Synthesize large-scale AMT dataset using Reaper and Native Instrument
collecting musicxml files from MuseScroe sheet music
converting musicxml files to midi format using batch converter plugin in MuseScore
saved as default (Maverick)
for the other two pianos, decrease the reverb volumn master to 5dB
write a python reaper project parser
Anaconda python and 32-bit python doesn't work, install 64-bit python, and add the python .dll file to Options->Preferences->ReaScript. Now, Reaper can recognize .py files in ReaScript!
To run the file, in Actions->Show Actions List load the created .py file, select it, and Run!
follow reapy instructions.
In the python which Reaper is referred to use (check which python), install reapy.
pip install python-reapy
check installation (within Reaper, run enable_dist_api.py and it works):
python -m reapy
test -> open Reaper, in shell, run
import reapy
reapy.print("hello world")
"C:\Program Files\REAPER (x64)\reaper.exe" -renderproject C:\Users\Marco\Downloads\test.rpp
For each piano model:
Save preferred FX chain as default.
Create an empty reaper projec with preferred render settings (here I am rendering to data/flac/{piano}/{project_name}.flac in flac format, 16 bit depth, mono). Save the template project to reaper_templates/template_{piano}.rpp.
run:
python build_reaper_projects.py <piano>
python render_flac_files.py <piano>
After synthesize all the piano models, run:
python move_references.py