Towards Image Design Space Exploration in Spreadsheets with LLM Formulae
J.D. Zamfirescu-Pereira, Shm Garanganao Almeda, Kyu Won Kim, Bjoern Hartmann · 2023
Users of Text-to-Image (TTI) models like DALL•E and Stable Diffusion typically engage in a lot of iteration, exploring a design space with two main inputs: (1) prompt text spanning image content and style; and (2) stochastic (e.g., random seeds) and other opaque (e.g., classifier-free guidance) variables. Here, we demo an early prototype interface using a spreadsheet metaphor to enable exploration and display of multiple input changes simultaneously, and affording prompt-crafting using spreadsheet formula construction. New LLM-based functions aid rapid exploration of the prompt text input space, by generating new variations on existing prompts and context-relevant lists of prompt keyword options.