Süner Pla Cerda S. (Executive), Ünlü C. E., Kumbasar E.
Project Supported by Higher Education Institutions, 2023 - 2025
Numerous Al-based text-to-image generation models have started to be commercially available in the past few years, making the end-user interfaces of such platforms widely accessible by the masses. The term text-to-image defines the computational methods which use human-written keywords or sentences to create images with related semantic meaning to the text. These Al models use a variety of methods for text encoding and image generation. Dall-E, Midjourney, and Stable Diffusion are a few such popular models to name a few. These tools enabled a much wider segment of the population, who are not necessarily trained as professional artists or designers, to create various forms of artwork regardless of their technical or artistic abilities. This sudden accessibility of fast and easy image production has brought about many discussions on creativity, authorship and ethica I conduct in art and design. Currently, public concerns are visible in the creative communities as to whether Al should be considered a threat or an opportunity in the arena of design. Pessimistic views suggest that, unlike past beliefs, Al has finally reached a point of threatening creative jobs. Others defend that Al can simply be considered as just another of the many tools used in the design process, aiding the designers with speed, efficiency and optimization. It is thus regarded as a means to elevate the human creative process rather than to replace it. The accessibility of Al-based technologies will undoubtedly transform the creative and production processes to the core. In order to understand the extent of their impact on professional practice, it is vital to investigate the pragmatic instrumentalisation of the DesignerAl collaboration in the design process, The goal of this project is to investigate the potential roles and limitations of commercially available text-to-image models in the design process. After a careful assessment based on preliminary hands-on explorations of many available tools, five Al-based text-to-image generation platforms were selected (DALL-E 2, Midjourney, Stability Al, Starry Al and Vizcom). In this project, these platforms will be tested with the prompt strategies synthesized from the literature, and the outputs wili be assessed by participatory workshops and online surveys. Three outputs are aimed at the end of the project: (1) a prompt guide to support various stages and activities of the design process and a complementary prompt-image coupling pool, (2) identifying the potentials and limitations of text-to-image platforms in supporting design at various design stages and activities such as metaphorical inspiration, form generation and iteration, visualisation and presentation, and (3) an extensive assessment of the attitudes of the design community towards the use of such platforms in design and their impacts on the profession.