Exploring the Role of Text-to-Image Ai in Concept Generation

Abstract

Artificial intelligence (AI) capable of generating images from a text prompt are becoming increasingly prevalent in society and design. The general public can use their computers and mobile devices to ask a complex text-to-image AI to create an image which is in some cases indistinguishable from that which a human could create using a computer graphics package. These images are shared on social media and have been used in the creation of art projects, documents and publications. This exploratory study aimed to identify if modern text-to-image AI (Midjourney, DALL-E 2, and Disco Diffusion) could be used to replace the designer in the concept generation stage of the design process. Teams of design students were asked to evaluate AI generated concepts from 15 to a final concept. The outcomes of this research are a first of its kind for the field of engineering design, in the identification of barriers in the use of current text-to-image AI for the purpose of engineering design. The discussion suggests how this can be overcome in the short term and what knowledge the research community needs to build to overcome these barriers in the long term.

Are generative design tools creative? A characterisation of tools throughout the design process

“…This process is illustrated in Figure 2. The 15 tools selected for review were: TRIZ (Gadd, 2011;Petrov, 2023), Functional Analysis (Little, Wood and McAdams, 1997), Brainstorming (Putman and Paulus, 2009), ChatGPT (Filippi, 2023), Morphological Analysis (Arciszewski, 2018), DALL-E (Bing, 2023;Brisco, Hay and Dhami, 2023), CATIA GDE (Dassault Systemes, 2020), Inspire (Lim and Wong, 2018), Ansys (Ansys, 2021), Abaqus (Johnsen, 2013), Sulis (Altair, 2023;Das et al, 2023), Siemens NX (Menninger, Berroth and Jacobs, 2022;Siemens, 2023a), AutoDesk Fusion 360 (AutoDesk, 2023), Simulink (MathWorks, 2023) and Amesim (Siemens, 2023b). The four italicised early-stage design tools in this list are not computational methods.…”

Inspiration or indication? Evaluating the qualities of design inspiration boards created using text to image generative AI

“…Specific to GenAI applied in inspiration activities for industrial designers, Mikkonen (2023) explore how inspiration boards (also known as mood boards and vision boards) can be supported via GenAI, however fall short from actually testing the approach. Brisco et al (2023) find from early explorations with image GenAI that designers see a clear potential for its use to support inspiration finding activities. At the same time, they highlight a current limitation of image GenAI alone to create viable/functional solutions.…”

“…Emerging research on the use image GenAI in design indicates a viable avenue to support inspiration via provision of inspirational image content (Koch (2017) that can be the basis of AI supported inspiration boards that stimulate design inspiration on the above-mentioned qualities of a product. (Brisco et al 2023;Mikkonen 2023). Inspiration boards are interesting as the subject matter in our exploration of image GenAI supporting inspiration as we contend they are a formal embodiment of inspirational material.…”

A New Approach to Interior Design: Generating Creative Interior Design Videos of Various Design Styles from Indoor Texture-Free 3D Models

“…Using generative AI models to produce images is a recent research hotspot [14,[18][19][20]. Their ability to generate images based on textual descriptions makes it possible to produce designs directly through texts [21][22][23][24]. However, two significant challenges emerge when applying generative models to interior design.…”