Image generation models respond to descriptive prompts. Clear descriptions produce better images.
Describe the Subject
What's in the image, with specific details: "a red vintage bicycle leaning against a stone wall covered in ivy".
Set the Style
Photography, watercolour, flat illustration, 3D render, pencil sketch. Reference general styles and eras rather than living artists.
Composition and Framing
Close-up, wide shot, overhead view, centred subject, rule of thirds.
Lighting and Mood
Golden hour, soft studio light, dramatic shadows, calm, energetic.
Details and Constraints
Colour palette, aspect ratio, background. Some tools accept negative prompts listing things to avoid.
Iterate
Generate several options, pick the closest, and refine the prompt. Use editing features to fix specific areas.
Common Weak Spots
Text in images, precise counts, hands and complex spatial relationships have improved but still fail at times.
Responsible Use
Respect rights and consent: avoid generating images of real people without permission, check licensing terms for commercial use, and label AI-generated imagery where appropriate.