How do I write prompts for AI image generation?
1 Answer
AI image generators turn text descriptions into pictures. The quality of the result depends heavily on the description. Here is a practical framework for writing better image prompts.
Include these elements:
1. Subject: who or what is in the image. Be specific. "An elderly man reading a newspaper" is better than "a man."
2. Action and details: what is happening, clothing, expressions and objects.
3. Setting: where it takes place, such as a busy railway station, a village courtyard or a modern office.
4. Composition and camera: close-up, wide shot, aerial view, eye level or low angle, and where the subject sits in the frame.
5. Lighting: soft morning light, golden hour, dramatic shadows, neon lights or bright studio lighting.
6. Mood and colours: calm, energetic, nostalgic, warm tones, pastel palette or black and white.
7. Style: realistic photograph, watercolour, oil painting, flat vector illustration, 3D render, pencil sketch, cartoon or minimalist poster.
8. Format: square, vertical for stories and reels, or wide for banners and thumbnails.
Here is an example:
Basic: "A tea stall."
Detailed: "A roadside tea stall in a small Indian town on a rainy evening, a vendor pouring chai from a height into small clay cups, steam rising, customers under a blue tarpaulin, warm light from a bulb, wet street reflecting the light, realistic photo, cinematic mood, wide shot."
Tips for better results:
Put the most important elements first.
Use concrete visual words rather than abstract ones. Instead of "beautiful," describe what makes it beautiful.
Keep any text in the image short, and put the exact words in quotation marks.
Describe one main idea per image; very complex scenes often go wrong.
Iterate. Change one thing at a time, such as lighting, style or angle, to learn what works.
Use editing. Many tools let you upload an image and request specific changes, which is often faster than regenerating from scratch.
Different tools respond differently. Some prefer natural sentences, while others work with comma-separated keywords. Conversational tools like ChatGPT and Gemini understand ordinary sentences very well.
Responsible prompting matters as well. Avoid prompts that depict real people in misleading or harmful situations, copy specific artists' work or brands, or create content that could deceive. Most tools block such requests anyway.
With practice, you will find that describing images precisely is a skill much like briefing a photographer or illustrator.