Question
What makes a good AI image prompt?
A good prompt describes the scene in full sentences: the subject, the setting, the light, the camera view and what the image is for. Google's guidance for its image models says that more detail gives you more control and that explaining the purpose helps the model.
Say what you want rather than what you don't. Google's example: instead of "no cars", write "an empty, deserted street with no signs of traffic".
For a busy scene, split the request into steps: the background first, then the main subject, then the details, as Google's guidance recommends.
Expect to refine. The first image shows what the model understood, and the next instruction corrects what it missed.
In ImageQuarry the refining happens in the same conversation: you ask for a change to the version you chose, and the earlier version is kept.
| Part of the prompt | Example |
|---|---|
| Subject | a chrome fragrance bottle |
| Setting | on folded ivory fabric |
| Light | hard afternoon sunlight from the left |
| Camera | close-up from a low angle |
| Purpose | a product page hero image with space for a headline on the right |
Source: Google's Gemini image generation best practices; checked October 20, 2026.
Related questions
Last reviewed October 20, 2026 by ImageQuarry
Create an image. Edit it with words. Keep refining.
Put a good prompt to work
Describe the subject, setting, light and purpose, then refine the result in the same conversation.
Start creatingAlready have an account? Sign in