A study of how the prioritization of prompts affects generative AI images : Focus on examples of generative AI works

Despite its unlimited potential and ability to create, generative AI is limited by its inability to accurately reflect human intent. This limits its usefulness, which is why a comparative experiment was conducted to improve its usefulness by performing planned tasks according to the creator's intention. This study is a case study of the hypothesis that prioritizing and arranging prompts can lead to results that clearly reflect the creator's intent. We used Midjourney, Gemini, and Co-Pilot to generate the images. The keywords were ranked on a 10-point scale, with the most important keywords at the front and the least important at the back, and their alignment with the researcher's topic. The success rate of prioritizing keywords was (single) 4.5 : 10 points for the keyword ranking experiment / (mixed) 4 : 10 points for the keyword + prompt ranking experiment / (single) 6.5 : 10 points for the prompt ranking experiment. The results confirm the hypothesis that the prioritization of prompts leads to clearer results, and that the information in the initial part of the prompt plays an important role in determining the clarity of the results. This study provides basic knowledge for creators and researchers who want to use generative AI.