This is a non-stable development prototype.
A musical composition generator capable of learning musical styles and generating compositions. Under the hood, Melodendron uses multiple viewpoints variable order markov models. Melodendron's design is inspired by Francois Pachet's Continuator.
- Implements a viewpoint agnostic multiple viewpoints variable order markov model tweaked for fast continuation generation.
- Midi file parsing and utilities are provided.
- Reduction functions library to enrich state's viewpoints.
- Selector functions library to control selection of continuations.
Parsing:
- Real-time parser.
- Legato tolerant parsing.
- Support for key and time signatures changes.
- Support for dynamic key detection.
Model:
- Selector function that weights viewpoints using entropy.
- Better metrics.
- Real time learning.
- Real time generation.
- Saving and loading of models.
See the examples
directory.