Gemini Image Editing Prompts: Fix AI Images Fast
Copy-ready Gemini prompts to edit an image you already have: swap the background, remove objects, fix text, go vertical, and what to do when it gets stuck.
- Published
- Reading time
- 5 min read
Contents
- The golden rule: say what stays the same
- Backgrounds and objects
- Text, light and format
- Style and combining images
- When Gemini gets stuck
- FAQ
- Can Gemini edit an image created with ChatGPT?
- Why does Gemini change parts of the image I didn't mention?
- Is Gemini free for image editing?
- Should I edit in the same chat or start a new one?
- Further reading
Generating a good image is half the job. The other half is the small fixes: a different background, one object too many, the wrong color, a version that fits a vertical video. Starting over in a generator wastes attempts and often loses what you liked about the first image.
Gemini is good at this second half. You upload the image you already have, describe one change, and keep everything else. This page has the editing prompts I use, plus what to do when Gemini stops cooperating.
Nil's note
My workflow is ChatGPT first, for the base image, then Gemini for the edits. Sometimes Gemini doesn't understand what I'm asking, or it gets stuck and returns the same image without changing anything. When that happens I either fix it by hand or wait until my ChatGPT limit resets and go back to it.
The golden rule: say what stays the same
An editing prompt has two parts, and most bad results come from leaving out the second one:
- The one change you want, as specific as possible: what, where, and how it should look.
- What must not change: the subject, the composition, the style, the colors, the text.
Asking for one change per message also helps. If you ask for three things at once and one goes wrong, you've lost all three.
Backgrounds and objects
Replace the background of this image with [new background, e.g. a plain light gray studio wall]. Keep [the subject, e.g. the laptop and the coffee mug] exactly the same: same position, size, lighting and colors. Match the shadows to the new background.
Naming the subject explicitly tells Gemini what to protect while it rebuilds everything behind it.
Remove the [object] from the [position, e.g. bottom right corner] of this image and fill the space naturally with the [surrounding surface, e.g. wooden desk]. Don't change anything else in the image.
Saying where the object is and what should fill the space avoids a smudge or a strange new object.
Change the color of the [element, e.g. hoodie the person is wearing] from [current color] to [new color]. Keep the fabric texture, the lighting and everything else in the image the same.
Pointing to one element by its position and current color keeps the change local.
Text, light and format
In this image, replace the text "[current text]" with "[new text]". Keep the same font style, size, color and position. Don't change anything else.
Quoting the exact old and new text is the clearest instruction you can give an image model.
Make this scene look like it's [time of day, e.g. late evening], with warm light coming from [direction, e.g. a desk lamp on the left]. Keep the composition, the objects and the style unchanged.
Lighting edits work best when you describe the light source, not just the mood.
Turn this image into a vertical 9:16 version. Keep the main subject centered and unchanged, and extend the scene naturally above and below it so nothing important is cut off.
Extending the image instead of cropping keeps the subject whole for TikTok, Shorts or Reels.
Style and combining images
Redraw this image in the style of [style, e.g. a flat vector illustration with thick outlines]. Keep the same layout, the same objects in the same positions and the same color palette.
Asking it to keep the layout means you get the same picture in a new look, not a new picture.
Take the [subject] from the first image and place it in the scene from the second image, [position, e.g. on the table in the foreground]. Match the lighting and perspective of the second image. Don't change the subject's shape or colors.
Uploading two images and naming each one removes the guesswork about what goes where.
When Gemini gets stuck
Sometimes an edit simply doesn't happen. These are the patterns I see and what usually fixes them:
| What happens | Why it usually happens | What to try |
|---|---|---|
| It returns the same image | The chat got long and the request is ambiguous | Start a new chat, upload the latest version, ask for one change |
| It changes everything | The prompt didn't say what to keep | Add "Keep everything else exactly the same" and name the subject |
| It ignores where you pointed | The position was vague | Use clear positions: "top left corner", "the mug on the right" |
| It refuses the edit | The request touches a real person, a brand or a policy limit | Rephrase without the person or brand, or do it by hand |
If two or three tries don't work, stop spending attempts. For small fixes, a manual edit in any image editor is often faster: removing a spot, adding text or cropping takes a minute. For bigger changes, go back to the generator you used for the base image with a fresh prompt.
FAQ
Can Gemini edit an image created with ChatGPT?
Yes. Upload the image file to Gemini and describe the change. It doesn't matter which tool created the original.
Why does Gemini change parts of the image I didn't mention?
Image models regenerate the picture rather than editing pixels like a photo editor. Telling it explicitly what must stay the same, and asking for one change at a time, keeps those side effects small.
Is Gemini free for image editing?
Gemini offers image editing on its free tier with usage limits that change over time. Check the current limits in the Gemini app if you plan to edit many images in one day.
Should I edit in the same chat or start a new one?
Use the same chat for a few quick changes. If results start getting worse or the image stops changing, start a new chat and upload the latest version.
Further reading
- Previous: ChatGPT image prompts that work
- Related: How to write better prompts for AI coding assistants
- Related: What are tokens in AI models?
Written by Nil Martinez
Developer · DAM & DAW student
Developer studying multiplatform (DAM) and web (DAW) application development. I work every day with TypeScript, Angular, React, Java, Python, FastAPI, MySQL and Docker on Kubuntu, and I write about the tools I actually use.
Keep reading
ChatGPT Image Prompts That Work: A Copy-Ready Library
Copy-ready ChatGPT image prompts for blog covers, thumbnails, product shots and icons, plus the prompt structure that makes them work and how to save free uses.
Prompt Library7 min read
What Are AI Agent Skills? Why They're So Useful
Skills teach Claude Code, Codex and Cursor to do a task your way. What they are, why they beat long prompts, and how they compare to plugins and MCP.
Skills & Plugins6 min read
How to Install Skills in Claude Code, Codex and Cursor
Where skill folders go in Claude Code, Codex and Cursor, how to install plugins and curated skills, the one command for every agent, and the Windows paths.
Skills & Plugins5 min read