I'd try experimenting going sideways with a different AI for some time. Try ChatGPT and Gemini and see how you get on with them. With Chat, you get about 2 images free per day. With Gemini, it's about 16 or more. Both of these respond to prompt specificity more tightly (they're LLMs, or "Large Language Models", and are "trained" on listening and response: Midjourney is highly sophisticated but its purpose is to create "arty" image responses to prompts and doesn't have capability outside of image generation). Chat is actually a go-between where the image output is more "arty" than you'll get from Gemini, but it's slightly susceptible to bad anatomy and extra fingers. Gemini can get anatomy wrong, but less than Chat; the tradeoff is that it's less artistically subtle. Also, how about logging into Meta AI? That's completely free, and a good testing ground for Midjourney, apparently (I haven't used MJ for years because of the pricing and I'm a cheap shot) - it uses a similar rendering engine to MJ. :) Do consider prompt length. My shortest prompt covers about 4 lines. Longest? You could probably fill a small book with it. 😅