How do I write a good prompt for AI image generators like Midjourney?

A good Midjourney prompt follows a simple structure — subject, setting, style, and technical parameters — and the more specific you are about the visual details, the closer you'll get to what's in your head. Beginners often type something like 'a cat in a forest' and wonder why the result looks generic. The AI isn't a mind reader. You need to describe the light, the mood, the camera angle, even the lens type. Here's a concrete before-and-after. Bad prompt: 'a wizard casting a spell.' Good prompt: 'elderly wizard with a braided silver beard, casting a glowing blue spell in a dark stone chamber, cinematic lighting from below, dust particles floating in the air, shot on 85mm lens, photorealistic, 8k.' The difference in output is night and day. I've tested this hundreds of times. One thing that trips people up is throwing in too many conflicting style words. If you say 'watercolor, oil painting, photorealistic, anime style,' the AI averages them into mush. Pick one visual lane and commit. Also, negative prompts matter. Tell the AI what you don't want — 'no blur, no extra fingers, no text' — and it'll clean up a lot of the weirdness. A tip most guides skip: study the prompts of images you like on Midjourney's public feed. You can reverse-engineer what works by seeing how experienced users structure their descriptions. **Related**: How many words should a good AI image prompt be? | Can I use ChatGPT to write Midjourney prompts for me?