AI Image Generator

Writing prompts

How to describe images that work as an X multi-image post, and how prompt improvement works.

A good prompt

Describe what the viewer should see, in plain sentences:

  • Subject: what is in the image (a coastal road, five lipstick tubes, a shelf of books).
  • Setting and light: where it is and how it is lit (golden hour, overcast, studio softbox).
  • Style or medium: photo, watercolor, 3D render, flat illustration.
  • Composition: what is centered, what fills the frame, what to leave out ("no text", "plain background").

Prompts can be up to 3,000 characters. Longer is not automatically better; a few clear sentences usually beat a list of keywords.

Prompts for a split panorama

Choose X Carousel Ready mode (see Generation settings) and a wide aspect ratio such as 16:9. Describe a scene that runs continuously from left to right, with the main subjects spread out rather than centered, so split lines fall on quiet areas. The mode guides the model; it can't guarantee that every slice lines up perfectly.

Prompts for a set of separate images

For a series (a color study, a collection), generate each image with the same structure and change one thing: the color, the object, the season. Keep style, light and framing words identical so the set looks consistent.

Improve prompt

Improve prompt asks a text model to rewrite your prompt with more visual detail. It costs 1 Credit:

  1. Select Improve prompt. A confirmation shows the cost.
  2. Confirm. The suggestion appears next to your prompt.
  3. Choose Use this prompt or Keep mine. Use my original prompt brings back what you wrote.

You are charged only when a suggestion is returned. If the request fails, nothing is charged. Repeating the same confirmed request (for example after a network retry) is charged once. See Prompt improvement.