Review article

The article presents substantiation for the effectivenessof using trained artificial intelligence to create architecturaland artistic solutions of buildings based on approbation ofexperimental design. The requirements for architecturalvisualizations are defined; the rationale for the choosingof the neural network used is substantiated together withdescribing of the process of creating a trained model,sketch preparation and image generation functions. Asfor a hypothesis and goal setting, there is proposed a new approach to creating architectural and artistic solutions of buildings using trainable models based on open neural network with an intuitive interface. It is shown that use of own trained models for generation of architectural images makes it possible to create solutions qualitative from artistic point of view which meet the specified conditions.