Magic Easel: Environment-Grounded Visual Inspiration for Gen AI Creativity Support

Abstract

GenAI creativity support tools often condition inspiration solely on user input, which can narrow exploration and contribute to fixation. We designed Magic Easel, a tangible creativity support tool that introduces the physical environment as an additional input source by detecting environmental objects and semantically matching them with users’ in-progress drawings to generate visual inspirations. We conducted a mixed-methods within-subjects study (N = 32) comparing three conditions: baseline, low-semantic-similarity inspiration, and high-semantic-similarity inspiration. Quantitative measures showed no significant differences in perceived usefulness. Qualitative analysis revealed that semantic similarity shaped interpretation modes: high-similarity encouraged shape-based reasoning and inhabited narratives, while low-similarity encouraged thematic reinterpretation and causal worldbuilding. Notably, environment-grounded inspirations increased divergent thinking relative to baseline, contrasting with prior work showing GenAI typically hinders divergent thinking. We contribute a novel system and empirical evidence that semantic similarity distance can influence how people interpret and incorporate opportunistic inspiration.