I started experimenting with a couple of different models for AI images (midjourney, Dalle3 and Adobe) but the results are pretty atrocious and often completely ignore the description I give. Do you have specific prompts you use or general advice on how to improve the results?