Image Prompt Designer

What this lesson is about

An image model will happily turn a vague prompt into something. The question is whether it turns your prompt into what you meant. This session builds a prompt in layers, deciding one thing at a time, and ends with a positive prompt, a negative prompt, and notes on the platform you are generating with.

What you'll be able to do

• I can say what makes an image prompt specific rather than vague, across six dimensions.
• I can order a prompt so the things that matter most carry the most weight.
• I can write a negative prompt, and explain what it is for.

A little background

Precision comes from being specific across six dimensions: subject, style, mood, lighting, composition and medium. "A person" becomes "a young woman in silver armour"; "cool" becomes "anime concept art"; "dark" becomes "tense and melancholic". Each one narrows the space the model is choosing from.

Order carries weight of its own. Image generators read a prompt top to bottom and give the most weight to what comes first, so the useful sequence is subject → setting → style → mood → lighting → composition → medium. Every prompt also has a second half: the negative prompt, listing what to keep out — blurry, watermark, extra limbs, distorted face, text. Using both gives far more control than a positive prompt alone. The check that your prompt is specific enough is whether someone else, reading it cold, would picture roughly what you pictured.