03

Conductor AI

The game audio employs two components -
 

• Generative Music: Music samples are generated by a Neural Network, offline, to be combined, modified and played in-game. The NN is based off Recurrent Neural Networks with LSTMs, that carry time context for playing different notes. The RNN is fed different kinds of samples corresponding to different needs/environments for the gameplay.
  
   

• Procedural Audio: Using in-game inputs like player health, stamina, etc, and game environment parameters like distance from the shore, enemies, etc, the generated-samples are altered and mixed together to give the effect that would be most desirable for the gameplay, at a given point in time. PureData, an open source visual programming language is integrated with Unity to create rich interactive music.