Repurpose a Photo You Already Have

Start with an example

Marcus sells hot sauce at Chicago farmers markets. He has one good photo: a bottle on his kitchen counter, with a dish rack and a phone charger in the background.

With that one photo and a few prompts, he creates:

• The same bottle on a rustic wooden table surrounded by fresh peppers, for his website
• A vertical version with clean space at the top for “New Flavor!”, for Instagram Stories
• A clean white-background version for his online store
• A flat illustrated version of the bottle for his market banner

One photo. Four uses. No photo shoot.

But when Marcus zoomed in on the first version, he noticed the AI had changed his label’s tiny ingredient text into gibberish. He fixed it, which is exactly the kind of check this lesson teaches.

Generating vs. editing

In Lessons 1 and 2, you generated new images from text. Now you’ll edit images you already have. Editing has two big advantages:

1. It keeps reality. Your real product, your real storefront, and your real team stay recognizable.
2. It’s more controllable. The AI starts with your photo instead of random noise, so the results stay closer to what you need.

The editing toolkit

Most modern image tools offer some or all of these techniques. The names vary by tool, but the ideas are the same.

1. Instruction editing

Upload a photo and describe the change in words: “Make it look like a snowy evening.” The tool decides what to change. This is fast but less precise.

2. Inpainting: edit a selected area

You brush over or select part of the image and describe what should go there. Everything outside the selection stays untouched.

• Remove a stain, a power line, or a stranger in the background
• Replace a plain wall with a brick wall
• Change the color of one object

Tip: Select slightly more than the object, so the AI can blend the edges naturally.

3. Outpainting: extend the edges

The AI generates new content beyond the original borders. This is how you turn a square photo into a wide banner or a tall Story without stretching or cropping.

Tip: Describe what should appear in the new space: “Extend to 16:9. Continue the wooden table and add more peppers on the left.”

4. Background replacement

Many tools can separate the subject from the background automatically, then let you place it somewhere new: a studio backdrop, a kitchen, or a seasonal scene.

5. Reference images

Some tools let you upload a photo as a reference for a new generation. The AI uses it as a guide for a product, a character, a face, or a style. This is the key to consistency: the same product or mascot across a dozen images. (AIH’s R.I.C.K. Bot videos are made this way.)

6. Style transfer

Turn a photo into a different medium: illustration, watercolor, or 3D. Useful for flyers, banners, and brand graphics.

Keep / Change / Remove / Output

Whatever the technique, structure your instructions the same way:

Part	Purpose	Example
Keep	Protect what must not change	“Keep the bottle and label exactly the same.”
Change	Describe the new look	“Place it on a rustic wooden table with fresh red peppers, soft morning light.”
Remove	Clear distractions	“Remove the dish rack and phone charger.”
Output	Set the size and finish	“Square 1:1, realistic, no added text.”

Being explicit about Keep is the most important part. Without it, the AI may “improve” things you wanted untouched, like a person’s face or your logo.

Check every detail

AI edits can change things you didn’t ask for. Zoom in to full size and check:

Check	What can go wrong
Labels and text	Letters become gibberish or slightly different words
Faces	Someone looks younger, older, or like a different person
Hands	Extra or merged fingers after editing nearby areas
Colors	A product becomes a slightly different shade
Edges	Blurry outlines, halos, or seams where new and old content meet
Repeats	Outpainted areas repeat the same pattern unnaturally

The honesty rule: If you sell it, the picture must match what customers receive. A sauce bottle that looks bigger or redder than the real one is false advertising.

Rights and permission

• Edit photos you own or have clear permission to use.
• Get permission before editing photos of other people, especially customers, employees, and children.
• Don’t remove watermarks or edit someone else’s work to pass it off as yours.
• Don’t alter images to mislead, such as editing a before-and-after photo or changing a news image.

A worked example: one storefront, five uses

Lena owns a flower shop in Logan Square. She has one daytime photo of her storefront. Here’s how she uses it:

1. Cleanup (inpainting): She selects a parked delivery van and a trash can, and prompts: “Replace with empty sidewalk and a potted plant.” She checks the edges: clean.
2. Seasonal versions (instruction edit): “Keep the storefront and sign exactly the same. Change the season to a snowy December evening with warm light glowing from the windows and string lights on the awning.” She makes a spring version too.
3. Website banner (outpainting): “Extend to 16:9. Continue the brick building and sidewalk on both sides.” She notices the outpainted bricks repeat oddly on the right, so she regenerates that side.
4. Story graphic (outpainting vertical): “Extend to 9:16. Add open sky above the building for a headline.”
5. Illustrated version (style transfer): “Turn this storefront into a charming watercolor illustration with soft edges, keeping the shop name readable.” The AI garbles the shop name, so she uses the illustration without the sign text and adds the name in Canva.

From one photo, Lena gets a full season of marketing images, and because she checked each one, all of them are accurate.

Think deeper: how AI edits without starting over

Editing a photo with AI uses the same process as making a new image, with one twist: instead of starting from pure static, the model starts from your photo with some static added.

A setting often called strength decides how much static. Low strength keeps your photo’s structure and changes only things like lighting or texture. High strength gives the AI more freedom, and your photo drifts further from the original.

Inpainting does this only inside the area you select (the mask), using the pixels around it as clues. That’s how it fills in the background where you removed something. Selecting a little past the object’s edges helps it blend. Outpainting does the same thing outward, extending the scene past the original frame.

Reference images work differently. The AI doesn’t edit the reference; it pulls out features like a face shape, a product’s look, or a style, and uses them to guide a new image. Because it’s interpreting rather than copying, small text, logos, and fine patterns often change. Always zoom in and check.

This is why Keep / Change / Remove works so well. The clearer you are about what must stay the same, the better the tool can protect it.

Questions to sit with:

• If a few clicks can remove a person from a photo or change a background, what does a photo actually prove anymore?
• Where’s the line between “touching up” a photo and changing what really happened?

Project

Pick one real photo you own. Make three new versions: a new background or setting, a cleanup that removes at least one distraction, and an outpainted resize for a different platform. Compare each to the original at full zoom and list every detail the AI changed that you didn’t ask for.

Using the photo I uploaded:
KEEP: [the person's face / the product and its label / the logo] exactly the same.
CHANGE: [the background to a bright kitchen / the lighting to soft daylight / the style to a flat illustration].
REMOVE: [the trash can on the left / the wires].
OUTPUT: [square 1:1 for Instagram / wide 16:9 for a website banner]. Realistic, no added text.