How to write AI image prompts

The prompt is the only thing a text to image AI sees. A clear prompt gets you a usable image on the first try; a vague one gets you something random. Use this formula.

The five-part prompt formula

1. Subject

Who or what is in the image: "a red fox", "a ceramic mug".

2. Setting

Where it is: "in a snowy forest", "on a marble counter".

3. Style

The medium: "watercolor", "3D render", "35mm film photo".

4. Light and mood

"Golden hour", "neon at night", "soft overcast", "cozy".

5. Framing

"Close-up", "wide shot", "top-down view".

Prompt examples

• "A red fox in a snowy forest, watercolor, soft morning light, wide shot"
• "A ceramic mug on a marble counter, product photo, studio lighting, close-up"
• "An old bookshop in Lisbon, oil painting, warm lamp light, cozy mood"

Common mistakes

Packing too many subjects in one prompt, using abstract words without visual meaning ("amazing", "perfect"), and changing everything between attempts. Change one detail at a time instead.

Ready to practice? Open the free text to image AI and try the examples above.