What are the best practices for building a text-to-image generation pipeline

Question

I am facing a problem in creating a system that can generate images based on text descriptions. I need help selecting the appropriate model and dataset, training the system efficiently, and dealing with potential challenges to ensure high-quality image generation.