Generation

How to construct a model for an utterance, then generate speech parameters from it.