How to write AI image prompts
A good AI image prompt says exactly what the picture should show. Below are the five parts our prompts are built from, and how to adjust one prompt for Gemini, ChatGPT and Midjourney.
The five parts of a good prompt
- The task: say what to do with your photo, for example "Transform my photo into…".
- What to keep: list the details that must stay recognisable: face, hairstyle, skin tone, expression.
- The subject and scene: describe the clothes, setting, props and action.
- The style: name the look you want: anime line art, studio portrait, 3D vinyl toy, film photo.
- Light, camera and format: give the lighting, the framing and the image shape.
A real example from our library
Here is the opening of our most-copied prompt, Anime Selfie Transformation:
Transform my photo into a high-quality anime portrait illustration. Keep my hairstyle, skin tone, expression and outfit details recognizable from the photo, but redraw me in crisp modern anime style — large expressive eyes with highlights, clean line art…
It covers the task, what to keep and the style in its first two sentences. That's why it works on most tools.
Adapting a prompt for each tool
Gemini
Upload your photo first, then paste the prompt as written. Gemini follows full sentences well. To change something afterwards, reply in the same chat ("make the background darker") instead of starting over.
ChatGPT
Attach your photo and paste the prompt. ChatGPT also reads plain sentences well. Add the shape you want at the end, such as "square image" or "tall portrait image". Ask for small changes one at a time.
Midjourney
Midjourney works best with shorter, comma-separated descriptions. Add your photo as an image reference and set the shape with a parameter. The anime example above could become:
crisp modern anime portrait illustration, large expressive eyes with highlights, clean line art, same hairstyle and outfit as reference --ar 4:5
Tips that help on every tool
- Use a clear, well-lit photo where your face is easy to see.
- Change one detail at a time, such as the colour, place or outfit, so you can see what made the difference.
- Results vary each time, so try the same prompt two or three times.