Welcome to SynthePG

Welcome,

This project is targetted for all designers / devs / searchers who love music and AI and team play.

This wiki page will explain you the genesis, aims and stated of the projet.

1. Genesis

This project started years ago when my young child refused to sleep... and after hours of nusing I started to think of a way to play simple but nice music to help him (me too?) to sleep.
After some chat with my friends, the idea to make this soft not only a single player but an instrument in a 'orchestra' emerged.

As a result the current projet status (fev 2018) is an Android app with jni linkage able to play 4 levels of music and producing partitions sheets in lilypond format.
The composer is a expert-system based on browninan noice for melody generation and random predefined rythms patterns selection.

1. Functions and caracteristics:
   Everything is synthetised, from samples to partiction in '.ly' format:

2. Note spectrum :

   • Notes are produced based on sin, square, triangle predefined signal.
   • An advance mode allow the user to define a temporal or frequential signal that is further used for sound synthesis,
   • multiple notes mixing,

3. Gammes and music style:

   • User can choose between pentatonic gamme, classical major/minor, chromatic AND old forgotten modal music (lyricrian, dorian, phrygian ...)
   • Also 'arabian' music style,

4. Harmonisation:

5. Melodic lead: a single melodic line with simple octave,

6. Full harmonisation: 4 singers music (soprano, alto, barython, basse) each with variing melody leading and basic harmonisation.

7. Aims and futur design

The aim of this projet is to update the original app on newer android SDK version and to bring a full new way of experience.
The recent appearance of TPG (tangled programs graphs) algorithm seems to feet the challenge of 'advanced' music composition.
Basically, the new app is intended to propose 2 modes.
3.1 A learning mode
Step by step, the TPG algorithm is proposing music samples and ask for a user felling.
Learning is done on rythms, melodie phrasing, harmonisation style, ...
A base of knowledge is kept in order to train the algorithm later.

3.2 A running mode
In this mode, the app is producing music based on the 'rules' learned before.
The produced music is strored in order to run training on it later.

Learning process:
As said before, the learning is intended to be steady and progressive in terms of scope (compositon, melodic line, arrangement, ornements, ...).
The expected behavior is the emergence of cohenrent nice music based on user preferences.