Text to Image System

A user study comparing text-based and choices-based input methods in generative AI image systems. We built Choices2I, a structured interface that lets users generate images by selecting visual attributes instead of writing prompts — and found that creative professionals preferred it regardless of expertise level.

Published through Chinese CHI 2024 and archived in ACM Digital Library (290+ downloads). The study proposed 5 design guidelines for building more adaptive AI creative tools.