Try simple prompts in Midjourney too. Yes, they have fancy parameters but I still find it takes time and light iterative tweaking to get the image that fits. Oh, but the twist is that many of the same prompts don't produce the same results. It's almost like GAI models are 'moody' and work great one moment and another time the same prompts produce completely unsatisfactory results. This is why I'm against reusing prompts but especially templating them.