Get from a vague image idea to the exact generation you pictured
The gap between what you imagined and what you got is almost always in the prompt's specificity.
- Outcome
- A prompt detailed enough to consistently produce what you actually pictured, refined through diagnosed iteration instead of random regeneration.
- Time
- 20-40 minutes
This workflow is part of PromptKit Pro
A handful of prompts and workflows stay free for everyone — Pro unlocks the full library, including this one, plus the guided runner.
Vague image prompts produce vague, inconsistent results, and the instinct to fix a miss is often to regenerate randomly rather than diagnose specifically what went wrong. This playbook builds specificity into the prompt from the start, then treats each miss as diagnostic information — what's different, and which part of the prompt is responsible — rather than a random reroll.
When not to use this
For quick, low-stakes generation where any reasonable result works, this level of process is unnecessary. This is for when you have a specific image in mind and keep missing it.
Did this work?
You converge on a prompt that reliably produces close to what you pictured, and can explain which specific words in the prompt are doing the work.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.