LUCID: Visual inspiration to editable geometry and artefacts

Our approach blends the intent-parsing and pattern reasoning strengths of large multimodal models with the precision of deterministic parametric design tools, with an emphasis on a transparent, human-editable provenance layer that links images, topology, and operations. The goal is to shift interaction from merely tuning parameters to actively co-authoring the generative process: branching, remixing, and selecting alternatives, while keeping a clear sense of how and why a design evolved and afterwards, fabricated.

Over the research of the project, we will prototype this mixed-initiative framework, evaluate how it expands design space exploration for design professionals, and test it through digital studies as well as selected physical prototypes, such as modular lighting elements that can stand alone or assemble into larger pieces. Results will be shared through a public showcase and accompanying research  dissemination.

The project is lead by Arbitrary Aesthetics, an AI CAX startup by Feng Ye & Janne Keskinen.

The project is part of art-ai-fact initiative.