There’s a lot of attention on image generation right now with the release of Google’s Nano Banana Pro (that’s Gemini 3 Pro Image Preview).
This new model, the “old” Nano Banana (aka Gemini 2.5 Flash Image), Open AI’s image models, and most of the other available image models will give you much better output if you spend some time on your prompts. There are a lot of elements that go into an image. If you leave things out of your prompts then the models will usually guess which might be what you want (but usually isn’t).
Here are some different elements to consider when prompting for images -
- Who or what is the subject of the image?
-
- Add details (i.e. a freshly groomed black German Shepherd wearing a blue top hat and a matching bow tie)
-
- Wide-angle, portrait, macro, etc.
- What is happening in the image?
-
- Juggling bowling pins with one hand and drinking an espresso with the other, jumping over the last hurdle before the finish line while barely in first place, etc.
- Where does the scene take place?
-
- Intergalactic coffee shop, strangely calm inside section of a hurricane, secret passageway of a mega-mansion,etc.
-
- Cyberpunk, pixel art, photorealistic, Sunday paper cartoon strip, etc.
- What is the aspect ratio?
-
- 9:16, 16:9, 1:1 square, 21:9 wide, etc.
-
- Golden hour, mid-day overhead sun, indoor halogen lamps, cinematic color grading with excessive red tones, etc.
- Camera details
- What kind of lens?
- What f-stop?
There’s so much more that could go into it. If you have any photographer friends, ask them what things they consider when setting up a shot. Just be ready for some words that you’ve never heard before. Or maybe you already know about bokeh…
Want another way to improve your AI-generated images? Watch this: