Think like a photographer
Image models respond best to visual language: what's in frame, where, how it's lit and what it was "shot" on. Abstract words like "amazing" or "beautiful" do very little.
- 1Subject: who or what, doing what
- 2Setting: where and when
- 3Lighting: golden hour, overcast, studio softbox, neon
- 4Camera: close-up, wide shot, 35mm, shallow depth of field, top-down
- 5Style: editorial photo, watercolour, 3D render, flat vector illustration
Example prompts
A barista pouring latte art in a small London café, early morning,
soft window light from the left, close-up, shallow depth of field,
editorial food photography --ar 4:5
Flat vector illustration of a person organising floating app icons,
pastel violet and sky blue palette, lots of white space,
clean geometric shapes --ar 16:9 --no textParameters worth knowing
--ar sets aspect ratio (16:9, 4:5, 9:16). --stylize controls how strongly Midjourney applies its own aesthetic. --chaos adds variety between the four results. --no removes elements. --seed helps reproduce a result. --sref uses an image as a style reference for consistency.
Tip: Use --no text rather than writing "no text" in the prompt. Mentioning a word can make it more likely to appear.
Using images commercially
Check your plan's terms before using images commercially, avoid prompting for real people, trademarks or a living artist's name, and disclose AI-generated imagery where your platform or client requires it.