A python toolkit to collect a dataset of sounds generated with the classic Synth1 VST.
If you are looking for the actual dataset, you can download it here
This toolkit is only useful if you want to generate a new dataset using presets that you have collected yourself.
Allows you to:
- Generate a neat CSV file with all parameters of several Synth1 presets
- Record a WAV file for each preset
Clone the repo and run
pip install .
If you are only interested in generating a CSV of the parameters, you can skip the next steps.
For the recording of the presets, the python code is only responsible for sending MIDI messages the the VST Host and recording the audio output.
I recommend using LMMS since it's open-source and super simple to use. I have only tested this code on Windows.
You will also need a way to route the MIDI signals for the Python process to LMMS. For this I used LoopMidi.
Create a new LMMS project and add a Track with the Synth1 VST
Set the MIDI input to the port created by LoopMidi.
Finally, load the midisettings.ccm in the Synth1 options. This tells Synth1 which MIDI CC to map to every parameter.
To generate a CSV containing all of the parameters, you need to collect all your Synth1 banks in a folder. Then
python -m synth1_dataset csv <path to banks_folder> <csv path>
To generate a WAV file for each preset you first need to generate the CSV. Then:
python -m synth1_dataset wav <path to csv> <output folder> --start <index to start from (default 0)>
You can find and edit some parameters in synth1-dataset/constants.py:
- NOTE_DURATION: The duration of each WAV in seconds.
- SOUND_DEVICE_NAME: The name of your soundcard. You can see the available soundcards by running:
python -m synth1_dataset devices - SOUND_DEVICE_HOSTAPI: The hostapi parameter of your soundcard (see command above).