A good AI image is mostly a good description. Not a longer one — a more specific one. This guide covers the structure that works, and what to do when a result is 90% right.
What you'll learn
- A four-part prompt structure you can reuse
- How to keep twenty images in the same visual style
- When to edit instead of regenerate
Step 1: Write the prompt in four parts
- Subject — what is in the frame
- Action or state — what it is doing
- Setting — where it is
- Style — how it should look
"A young cook plating a dish, in a small warm kitchen at night, flat editorial illustration with soft shadows" beats a paragraph of adjectives every time.
Step 2: Lock your style
Write one style sentence and paste it, unchanged, at the end of every prompt in the set. Changing even a word between prompts is the most common reason a set stops looking like a set.
Keep your style sentences in a notes file per project. A channel usually needs two or three total, not one per video.
Step 3: Edit the near-misses
If the composition is right and one element is wrong, use AI Image Editor. Regenerating gives you a different image; editing gives you the image you wanted. That distinction is worth a lot of time over a full project.
Common mistakes
- Stacking adjectives instead of naming the subject clearly
- Asking for text inside the image rather than adding it in Thumbnail Maker
- Regenerating twenty times when one edit would have finished it
Want to try this yourself? Generate your first set in 7LabsAI.
Try 7LabsAI