Let Images Define Nouns and Prompts Define Verbs
Let images define stable nouns and prompts define the remaining verbs.
Capture
After reference images were introduced, prompts continued to repeat the appearance of the woman, child, clothing, kitchen, violin, and furniture.
The prompts became longer without becoming more reliable. Slight differences between the text and image sometimes altered details that had already been approved.
Why
Repeating important information appeared to reinforce it.
Agreement between the reference and prompt seemed safer than relying on either one alone.
Why-Not
The image had already fixed the nouns.
Restating them required the generator to reconcile two descriptions of the same visual facts. Perfect agreement added little; imperfect agreement reopened settled decisions.
The scar was learning that redundancy can become conflict when two control surfaces describe the same thing.
Commit
Decision: Let reference images define the visual nouns of the scene and let the prompt primarily define verbs.
Use imagery to hold people, props, environments, wardrobe, and composition. Use text to direct actions, interactions, timing, objectives, and camera behavior.
Do not restate fixed visual details unless they are intended to change.
The reference image is not only an identity container, however. Its pose and composition also imply motion. The next decision accounts for that.
Timestamp
2026-07-24