Developing Flask based Image Generator Application using OpenAI’s DALL-E and Replit: A Comparative Analysis

This study dwells into the development of a web application which generates images based on a creative prompt (text input) from user and other various inputs according to user’s wishes and requirements. The images have been generated based on user's descriptions and inputs using OpenAI’s DALL-E image generation model through an OpenAI API Key. Python Programming language has been used, especially Flask which is an effective backend framework of the dynamic language. For UI and frontend, HTML/CSS and JavaScript were leveraged to provide the user with a smooth and seamless experience. Replit, an IDE has been used for the purpose of development, management and hosting. The results of the images generated from OpenAI’s DALL-E 2 and DALL-E 3 in our application have been further compared with images generated from models developed by other tech giants (Meta, Stability AI) using three metrics – Sharpness, Entropy and Inception Score. Neural Style Transfer was leveraged to discover those factors which might have been considered by OpenAI for improving DALL-E. This study includes application of Image Processing, Computer Vision and CNNs.