Prompt Optimization for generative art models
This paper discusses the optimization of prompts for generating AI art [1]. Popular AI models, such as stable diffusion and MidJourney, are explored by guiding image generation with prompts. However, the generated art’s quality depends on the prompts’ quality, which may not always reflect the user’s desires. To address this issue, we present a novel approach to improve user prompts through epimerizing, using a trained machine learning model on corpus data. Our approach is demonstrated to improve the quality of AI art generated with prompts, providing a more accurate reflection of user preferences. This research contributes to AI art generation, highlighting the importance of prompt optimization for achieving high-quality results. Index Terms—Prompt optimization [2], text-to-image generation [3], AI art generation [4], MidJourney [5], ML, NLP [6], Prompt Engineering [7]