Ideas for AI-generated image generation using diffusion models, with a comparative study and the impact of command-line engineering

Abstract

This research aims to study image generation techniques using artificial intelligence, focusing on diffusion models, and to analyze the performance of the three leading tools: DALL•E, Midjourney, and Stable Diffusion.

The research also examines the impact of prompt engineering on the quality of the resulting images.

Two practical experiments were conducted. The first compared the three tools using the same prompts within different styles (realistic, surreal, and Arabic). The second experiment focused on analyzing the impact of prompt engineering quality (weak, medium, and strong) using the same tool.

The results showed that Midjourney excelled in visual quality and artistic creativity, while DALL•E demonstrated accuracy in executing prompts. Stable Diffusion, on the other hand, offered high flexibility, although the quality of the resulting images varied. The study also proved that improving prompt engineering leads to a significant improvement in the quality of the resulting images.