NEURAL NETWORKS IN ART AS A GRAPHIC DESIGN TOOL

Abstract

The article examines the possibilities of various neural networks as tools for the implementation of projects in graphic design, evaluates their ability to ensure quality and efficiency in the creation of visual content for various types of products. The advantages and disadvantages of each neural network are also analyzed. The work presents the opinions of scientists and practitioners about the variety of neural networks that can be used to perform graphic design tasks. In addition, the results of own practical experience of working with neural networks are given. The study confirmed the effectiveness of neural networks in creating concepts of characters and locations for computer games, illustrations for printed and electronic publications, as well as in the development of trademarks and logos, corporate style and graphic design of packaging. However, their functionality does not yet provide the necessary quality level for such products as posters created on the basis of figurative language tropes; fonts; engineering graphics in axonometric projections showing the internal structure of devices or equipment; layout for print publications, websites and mobile applications, as well as infographics based on stylized images and design solutions for packaging. Maze Guru, Midjourney, and Leonardo AI are best for graphic design content. The ChatGPT neural network is an effective tool for matching peers and gathering feedback from scientists. The advantage of using neural networks is a significant acceleration of the process of creating visual content, as well as the possibility of combining different programs to supplement and improve the results obtained. Disadvantages include mainly English-language communication between the user and the network, as well as discrepancies between the images that exist in the user's mind and those generated by the network. Creations created by neural networks are easily recognizable, and for similar text queries, they can give very similar results.