Text-to-image generators, fueled by deep learning, are transforming the AI art landscape by enabling users to create artwork through text prompts without requiring design skills. This article explores the use of such generators and delves into a dataset of over 200,000 prompts from Midjourney users, available on HuggingFace. By using Python and the Cohere platform, the article guides readers through processing these prompts with word embeddings to explore similar prompts and cluster them into topics using unsupervised techniques like KMeans clustering. The process includes visualizing these clusters with scatter plots and word clouds, emphasizing the importance of crafting detailed prompts for higher fidelity image generation. The article also demonstrates how semantic search can enhance prompt creation by calculating similarity scores, offering insights into optimizing AI-generated art. Future tutorials are hinted at, promising the development of a prompt generator that could further streamline the creative process.