Abstract

Creative design could benefit from recent advances in AI, by providing inspirational images generation. While researchers have studied general benefits and limitations of existing text-to-image tools, this work investigates how to overcome these limitations and enable non-disruptive multimodal prompting for project-appropriate human-AI conversation. This paper presents the development of a system for human-AI co-creativity that supports text-based and sketch-based image generation through the elaboration of a complex prompt structure. The developed interface supports dialog interaction and the visualization of the respective contributions. Furthermore, this system constitutes a means for further studying human-AI interaction and how it varies in regard to conversation modality, between text and sketch.