The ArtificialSongGenerator tool was used to compile the open research dataset Artificial Audio Multitracks (AAM), that currently consists of 3000 artificial songs with rich annotations.
Please find more details on the project in the accompanying paper, that was published open-access in EURASIP Journal on Audio, Speech, and Music Processing.
For creating MIDI, download the jar file from the latest release and run:
java -jar AtificialSongGenerator.jar
The tool will generate a bunch of midi files based on the standard configuration.
You can specify your own settings by running:
java -jar ArtificialSongGenerator.jar --create-dummy
The config dummy file dummy.conf will then be created, which you can modify and feed back in to the generator by running:
java -jar ArtificialSongGenerator.jar --config=dummy.conf
All options and full usage info is provided by:
java -jar ArtificialSongGenerator.jar --help
The main directory provides an ant build file (build.xml).
Use ant pack to generate an executable jar (including downloading the JFugue library).
The main build file will call the build file of the subproject ArtificialSongGenerator/.
Then use bash script Sampling/generate-songs under Linux to create a whole dataset of songs based on the ArtificialSongGenerator routine.
Using Ardour or Mixbus one can then create audio using the Lua script Sampling/mixbus_roboter.lua therein.
Currently, the routine lacks full automization. This is planned to be accomplished next.
For more usage information:
All jars and scripts have usage and help files included via --help
In case you need any more help, find a bug or want to raise an issue, don't hesitate to contact me!