Prompt Programming for the Visual Domain

Alayt Issak, Lav R. Varshney

Keywords: Prompt programming, Visual Prompting, Prompt Engineering

TL;DR: Creativity in AI

Abstract: In this work, we ask how text-to-image synthesis via large language models can effectively probe imagery that embodies fidelity and imagination. We investigate this question in the context of prompts (writing to language models) in a novel probing mechanism known as prompt programming, or programming in natural language. We start by refining existing techniques to characterize the effect of templates on visual fidelity then hone in on approaches to capture holistic nuances within the visual domain. We present a systematic analysis of prompt engineering for visual image generation.