Personality Generation is capable of controlling different styles of personality and feelings, without losing fluency and while maintaining versatility that in any combination of differentiable models contributes to the generation of text. The model incorporates a pre-trained language model with one or more attribute classifiers that direct the generation of new text. These attribute models are small and easy to construct: a bag of words or a representation that defines the occurrence of words in a text snippet; and a linear discriminator, or a system that characterizes two or more groups of objects or events.