How to write image prompts that actually work

Most weak images come from vague prompts. The model fills every gap with its own defaults, which is why so many AI images look the same. The fix is not longer prompts, it is more specific ones.

Use a five part structure

• Subject: who or what, with one or two concrete details. "A ceramic coffee mug with a chipped rim" beats "a mug".
• Setting: where it is and what surrounds it.
• Light: time of day, direction and quality. "Soft window light from the left" changes everything.
• Lens and framing: close-up, wide shot, 85mm portrait, top-down flat lay.
• Style: photographic, film stock, illustration, 3D render. Pick one.

Before and after

Before: "a cozy cafe".

After: "A small corner cafe at 8am, steam rising from a flat white on a marble counter, soft window light from the left, shallow depth of field, 35mm film photo, warm muted colors".

Cut words that do nothing

Words like "beautiful", "amazing" and "high quality" rarely help. Describe what makes it beautiful instead: the light, the material, the composition.

Iterate one variable at a time

When a result is close, change one thing and generate again. Changing the light and the lens and the style at once makes it impossible to learn what worked.