The whole anatomy in one line
Every prompt on this site follows the same skeleton. Learn it once and every wall, guide, and tool here will make sense.
The subject is yours. The style and lighting words come from the walls. The details are the two things the image would be wrong without. That is the entire theory. The rest is picking better words.
Step 1: write the subject plainly
Say what is in the picture and what it is doing. No mood words, no "beautiful", no "epic". Those come later or never.
If you cannot name a subject, press the dice in the builder and argue with what it deals you. Disagreement is a fine way to find out what you want.
Step 2: take four words from two walls
Open the styles wall and click one word. Open the lighting wall and click one more. Style plus lighting moves an image further than any other pair. Add two details of your own.
The tray at the bottom of every wall collects your clicks and copies the line out. The builder does the same thing with dice.
Step 3: paste it where it goes
Midjourney: type /imagine in Discord and paste. The tray's Midjourney toggle formats the line for you. Stable Diffusion, Nano Banana, and GPT-Image: paste the plain line, no slash, or wrap it in a sentence for the chat image models. The Midjourney and Stable Diffusion pages explain the parameters when you want them.
Step 4: change exactly one thing
The picture came back wrong somewhere. Good. Swap the style word and generate again. Keep everything else identical. Now you know what that word does, and that lesson is permanent.
Change two things at once and you learn nothing. This is the one habit that separates people who prompt well from people who reroll and hope.
Where to go next
- Print the one-page cheat sheet: the skeleton, the parameters, ten words per wall.
- Browse the fifteen walls. Learn mode (📖) explains any word you tap.
- Let the builder deal words when the blank box stares back.
- Read Prompt Engineering, Properly when you want the why behind the skeleton.
- Use a chat model as your prompt writer with the image prompt guide.
Published here first.