# Change the scene. Keep the product.

One reference-led edit and a product-accuracy review.

## What you need

An owned product photo, approved product facts and an image-editing tool. Check access and cost before generating. The course's fictional mug reference is a practice input, not an Images 2.5 output.

## The steps

1. List five features that identify the product. Mark anything the photo cannot establish as unknown. Photograph another angle if the requested view would require inventing a hidden surface.

2. Attach the reference and fill the prompt. Request only a scene change. Save the first result even if it fails; record the exact model or app label, elapsed time and any displayed charge.

3. Compare the original and edit at full size: silhouette, openings, label, material and scale cues. Allow one repair, naming only the failed detail. Stop if it still changes the product.

4. Share the reference, attempt and checks in Content machine. Mark fictional practice products clearly. A generated image is not evidence for a product claim.

## Prompt

```text
Edit the attached product photo.
Product identity to preserve: [five visible details].
Change only: the background to [scene] and its matching contact shadow.
Keep the product's outline, proportions, markings, material and camera angle unchanged. No new accessories, labels or features.
Leave room at [position] for text I will add in an editor.
If the scene requires showing a surface absent from the reference, keep the original view instead of inventing it.
```

## Review

- [ ] All five identity details still match the reference.
- [ ] No unsupported product feature or claim appears.
- [ ] The review log includes rejected attempts, model label, time and cost if shown.

## If it fails

Return to the original reference, not the distorted result. Name the incorrect feature and request the same scene without changing that feature. Composite the real product in an editor if fidelity remains insufficient.

## Recall later

Which matters more for a product edit: a prettier room or an unchanged silhouette?

Product identity is the acceptance gate. A better background cannot compensate for a different product.

Tool: https://chatgpt.com/
Written by The Leverage Club. Reference edition 2026-09-08.

## Evidence and limits

Original first-look exercise, not a club benchmark. Earlier course images are practice inputs, not Images 2.5 outputs. Check current access and cost.

Launch: https://openai.com/index/introducing-chatgpt-images-2-5/
API guide: https://developers.openai.com/api/docs/guides/image-generation

## Your run log

Date:
App and exact model label:
Input filenames and permission:
Requested change:
Accepted or rejected, and why:
Elapsed time:
Displayed cost (or unknown):
Next action: