I was experimenting with AI image generation for a case study visual.
The output was close, but not exactly what I wanted.
I kept refining the prompt:
Change the color theme.
Adjust the needle position.
Match the layout better.
Make it feel more realistic.
Each iteration improved something, but the AI was still missing the exact visual structure I had in mind.
I created a quick flow sketch to show the direction more clearly:
Initial prompt → AI output → refined prompt → rough sketch/wireframe → final output
This helped me see where the gap was. The issue was not just the prompt; it was the missing visual context.
Then I tried a simpler approach.
I gave it a rough sketch.
Not a polished design.
Not a perfect wireframe.
Just a quick visual cue showing placement, spacing, and layout.
And that worked better than adding more words.
The sketch helped communicate the part that text was struggling to explain:
where things should go and how they should relate to each other.
That was the small learning for me.
Prompting is not always about writing longer instructions.
Sometimes it is about giving AI the right kind of input.
For visual work, quick sketches can help with:
Layout — where elements should sit
Spacing — how much room each part needs
Hierarchy — what should stand out first
Composition — how everything should feel together
When AI does not understand your visual intent, try sketching it first.
It does not need to look good.
It just needs to make your direction clearer.
Have you tried using quick sketches or wireframes to guide AI outputs?

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.