Refine an image prompt

What it asks you for

The idearequired

What you want a picture of.

Where it goes

So the framing and format fit.

What it tells the AI to do

Turn this into a prompt that renders what you meant.

Idea: {{ idea }}
Use: {{ usage | default("not specified") }}

A weak image prompt is a noun. A strong one names the subject, the framing, the
light, the medium, and what is NOT in it.

Work through, in this order: subject and what it is doing; composition and
crop; lighting and time of day; medium or style, named specifically rather than
as "high quality"; colour; and what to exclude.

Cut every word that is not doing work. "Beautiful", "stunning", "8k",
"masterpiece" steer nothing and dilute the words that do.

Return the refined prompt, then one line on what you decided that the original
left open, so the user can correct that choice rather than the whole prompt.

Why this works

A weak image prompt is a noun. The words that actually steer a model are the ones naming framing, light, medium and what to exclude, and the words that steer nothing are the ones people add most: "beautiful", "stunning", "8k", "masterpiece". This adds the first kind and cuts the second, then tells you which choice it made for you so you can correct that rather than start over.

• image
• prompt
• refine
• generate

More images skills

• Create an imageGenerate an image from a text prompt with a chosen model.
• Edit an imageEdit an existing image using a natural-language instruction.
• Remove the backgroundCut the subject out of an image and deliver transparent, white, and brand-colour versions.
• Upscale an imageEnlarge an image without re-generating it, preserving detail.
• Resize for every formatTurn one master image into every aspect ratio you need, cropped so the subject survives.
• Product shotPlace a real product into a new scene without altering the product itself.