Inkspire: Supporting Design Exploration with Generative AI through Analogical Sketching

TL;DR

Inkspire addresses design fixation in generative AI-assisted design by introducing a sketch-driven workflow that combines analogical inspiration with a stroke-by-stroke design generation loop and sketch scaffolding. The system integrates an Analogy Panel, Sketch2Design, and Design2Sketch to ground abstract concepts in visually concrete anchors and to convert AI outputs into low-fidelity scaffolds that support iterative exploration. In a within-subjects study against a ControlNet baseline, designers using Inkspire reported significantly higher inspiration and exploration, and improved perceived collaboration with AI across controllability, communication, and partnership. The findings suggest that analogical grounding and per-stroke generation foster a more co-creative, stateful interaction with GenAI, enabling broader design exploration while maintaining design quality and user satisfaction, with potential to generalize to additional domains and tasks.

Abstract

With recent advancements in the capabilities of Text-to-Image (T2I) AI models, product designers have begun experimenting with them in their work. However, T2I models struggle to interpret abstract language and the current user experience of T2I tools can induce design fixation rather than a more iterative, exploratory process. To address these challenges, we developed Inkspire, a sketch-driven tool that supports designers in prototyping product design concepts with analogical inspirations and a complete sketch-to-design-to-sketch feedback loop. To inform the design of Inkspire, we conducted an exchange session with designers and distilled design goals for improving T2I interactions. In a within-subjects study comparing Inkspire to ControlNet, we found that Inkspire supported designers with more inspiration and exploration of design ideas, and improved aspects of the co-creative process by allowing designers to effectively grasp the current state of the AI to guide it towards novel design intentions.

Figures (13)

• Figure 1: Sketch2Design pipeline, including (a) inspiration generation with LLMs, (b) sketch-guided design generation, and (c) foreground extraction.
• Figure 2: Comparison of our Design2Sketch method with potential alternative methods, such as edge detection, manga line extraction models, and models trained explicitly on pairs of sketches and images.
• Figure 3: Design2Sketch pipeline, including (a) semantic segmentation, (b) soft edge extraction, and (c) computing an intersection.
• Figure 4: Results on creativity measured with the Creativity Support Index (CSI) cherry2014quantifying (7-point Likert scale, higher is better).
• Figure 5: Example designs created by participants using Inkspire for the design tasks of designing a fluid chair and a serene lamp. The final participant-generated sketch is shown on the top, and the generated T2I image is shown on the bottom, along with the selected analogy word chosen by the participant.