Question
How should I phrase feedback to an AI image tool?
Name one concrete change, where it is, and what must stay the same, for example: "Make the light warmer from the left. Keep the face, crop and background." Avoid judgments like "make it better", which give the model nothing specific to change.
Good feedback reads like a note to a retoucher: a visible thing, a direction and a limit. "Move the bottle to the right third and keep the shadow" works; "it feels off" does not.
Google's documentation describes multi-turn conversation as the recommended way to iterate on images, which suits short, specific notes one after another better than one long list.
Useful verbs: replace, remove, add, move, warm up, cool down, crop, widen. Useful limits: keep the face, keep the product, keep the crop, keep the background. If a note is ignored, repeat it alone rather than adding more.
ImageQuarry's own example of a request is "Use this texture on the jacket. Keep the face, hands, crop, and background." Each note like that produces a new version; review it against the last one to see what changed.
| Instead of | Try |
|---|---|
| "Make it better" | "Make the light softer and warmer. Keep the product and the background." |
| "More premium" | "Replace the plastic table with dark marble. Keep the product." |
| "Fix the background" | "Replace the background with a plain warm gray wall. Keep the person and the crop." |
Sources: Google's Gemini image generation guide (multi-turn editing), checked October 20, 2026; ImageQuarry homepage (example request).
Related questions
Last reviewed October 20, 2026 by ImageQuarry
Create an image. Edit it with words. Keep refining.
Say it the way you'd tell a retoucher
Give ImageQuarry one clear note and see the result as a new version.
Start creatingAlready have an account? Sign in