Image prompts are changing.
The old style was often a long list of visual words. The newer workflow is closer to giving art direction and then revising.
Start with the job of the image
Before style words, decide what the image must do.
Is it an eyecatch? A product image? A diagram? A thumbnail?
The purpose changes the prompt.
Add concrete constraints
Good constraints help more than decorative words.
Use details like aspect ratio, subject, background, safe area, no text, or where the main object should sit.
Revise instead of starting over
A better image workflow is iterative.
First make the image. Then fix one problem at a time: crop, subject size, color, text space, or realism.
My rule
A good image prompt says what the image is for, what must be visible, and what must not appear.
Original Japanese Article
This article is based on the Japanese post: image2.0で何が変わった？ChatGPT Images 2.0時代のプロンプトの書き方.
