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

A good Midjourney prompt has three layers: the subject, the style, and the technical parameters — and you need all three to get predictable results. Beginners often stop at the subject. They type 'a cat wearing a hat' and wonder why the output looks like a blurry watercolor from 2004. The AI isn't a mind reader. If you don't specify a style, it defaults to whatever its training data suggests is average, which is rarely what you want. A stronger prompt looks like this: 'A ginger cat wearing a tiny blue top hat, sitting on a velvet cushion, soft window light, 35mm photography style, shallow depth of field, --ar 3:2 --v 6.' See the difference? You've told it what to draw (subject), how it should look (photography style, lighting), and how to format it (aspect ratio, model version). The technical bits at the end — those parameters starting with '--' — are where you control aspect ratio, stylization strength, and chaos levels. A tip most tutorials skip: use the '--no' parameter to exclude things you don't want. If your cat keeps ending up with a creepy human smile, add '--no teeth, human features.' It's surprisingly effective at steering the output away from common AI weirdness. You'll still get some duds. That's normal. But your hit rate — images you'd actually use — will climb from maybe 1-in-10 to 1-in-3 with this layered approach. **Related**: What's the difference between Midjourney and DALL-E 3 for beginners? | How many images do I need to generate before I get a usable one?