Emi: UI Components for Generating Emotional Images

Text-to-image generators have transformed the creative landscape by converting text descriptions into unique visual representations. However, text as an input modality presents challenges for users and artificial intelligence (AI) models due to its vast possibilities and the need for precise instructions. Conveying emotions through text-to-image prompts is particularly challenging, and research that has addressed user needs and the usability of such systems has been limited. Thus, we explored which user interface (UI) components could be useful for generating emotional imagery for mood boards. A co-design workshop was held to identify aspects that are important for emotional imagery and to gather ideas on how UI components for text-to-image generators could incorporate these aspects. Based on the findings from this workshop, a clickable prototype for a structured search, named Emi, was developed and evaluated. Our evaluation shows that Emi’s individual UI components are very useful in facilitating the creation of emotional images, and Emi as a whole is user-friendly, balancing both functionality and aesthetics well.