If you have ever typed a vague prompt into an AI image generator and got something random back, the fix is usually structure, not luck. Here is the template I use for predictable results.
The template
[subject], [setting], [lighting], [camera / composition], [style], [color palette]
Example:
ceramic coffee mug, minimal wooden desk, soft morning window light, top-down shot, clean product photography, warm beige tones
Why each part matters
- Subject first, so the model anchors on the main object.
- Setting gives context and avoids blank or random backgrounds.
- Lighting has the biggest effect on realism: "soft window light", "studio softbox", "golden hour".
- Camera / composition controls framing: "close-up", "top-down", "wide angle", "rule of thirds".
- Style sets the look: "product photography", "flat illustration", "watercolor".
- Color palette keeps a series consistent.
Tips
- Change one slot at a time when iterating, so you know what caused the difference.
- Keep a small prompt library in a text file and reuse the structure across a campaign.
- Keep text on the image short; most models still struggle with long lettering.
- For motion, the same structure works for image-to-video prompts if you add a camera move such as "slow push-in".
I tested this template with Freepik AI, which also handles image-to-video and text to speech, but the structure works with any image model.
What slots would you add to the template?
Top comments (0)