Free accounts in the US receive 120 credits per day; daily allowances vary in other regions. You can start exploring image ideas with Seedream 5.0 Lite and Nano Banana. Dreamina can help you change a product photo's setting or make a targeted visual revision. AI can also change details you intended to keep: a label, seam, reflection, handle, proportion or material. The risk depends on how much of the image is being regenerated and how closely the output is reviewed against the real item.
For a product listing, start with an accurate photograph and preserve the photographed item wherever possible. Use generation for the surroundings or a clearly defined creative change. Compare the final image with the original before publishing, even when the prompt says “keep the product unchanged.”
The key distinction is between a convincing image and an accurate representation of the product. A model can produce the first while missing the second.
- Which edits are most likely to change the product?
- Why “keep everything else unchanged” does not settle it
- Start with the product's identity, not a generic quality score
- A targeted Dreamina edit: visually preserved does not mean pixel-identical
- Choose the editing route that fits the required accuracy
- What can a dedicated product-repair tool do?
- Check the export in the place it will be used
- When should you stop repairing and take another photograph?
- FAQ
Which edits are most likely to change the product?
A full scene reconstruction gives the model more freedom than placing an original cutout on a new background. A local instruction narrows the intended change, but its actual effect still needs checking.
These are practical differences in the editing task, not a numerical failure-rate ranking.
Why “keep everything else unchanged” does not settle it
A prompt describes what you want.
An image-generation workflow may reconstruct the scene to satisfy the instruction. The result can preserve the overall silhouette while changing a small label or texture. A model may also fill an obscured area with something plausible rather than something that exists on the real product.
This matters most when the details affect the purchase: a bag's closure, a garment's pattern, a device's controls, a bottle's capacity marking or a material's actual finish. Do not judge those details only by whether the image looks professionally photographed.
A retained original layer provides more direct control over the visible product than a fresh reconstruction. It still needs edge and compositing review, especially for glass, reflective metal and fine fabric.
Start with the product's identity, not a generic quality score
Before editing, write down the features that distinguish the actual item. For a tote, that might include handle color, pocket outline, seam placement and a snap. For a bottle, it could be label wording, cap shape, fill appearance and the relationship between the label and glass body.
Keep this list short enough to use. It should describe the real product, rather than a vague instruction to preserve “all details.”
Review in three passes:
- 1
- Identity: Is this the same item, variant and construction? 2
- Information: Are labels, quantities, text and visible features accurate? 3
- Presentation: Do lighting, background, scale and shadow fit the intended image?
That order helps prevent a beautiful new setting from distracting you from the wrong item.
For a full tool decision, our AI product-photo comparison distinguishes generation, isolation and catalog-oriented workflows.
A targeted Dreamina edit: visually preserved does not mean pixel-identical
We used a synthetic cream tote with navy handles and a navy front pocket. The intended change was narrow: make the pocket warm terracotta while keeping the rest of the bag recognizable.
The Dreamina output was 1792 × 2240. In the native image and a smaller visual comparison, the pocket changed color while the handles remained navy. The pocket outline, stitching, snap, cream body, bag shape, background and overall lighting were visually retained.
The terracotta pocket gives the tote a fresh color direction while the indigo handles keep a familiar contrast. This is useful for exploring a concept variant before bringing the selected design into a campaign layout.
There is another commercial distinction: if the real product does not exist with that terracotta pocket, this image should not be presented as a photograph of an available variant. A technically successful recolor still needs accurate product information behind it.
Choose the editing route that fits the required accuracy
Keep the original product for an exact listing image
If the photographed item already looks correct, preserve it and change the surroundings. An original-photo panel is often enough for an ad; a carefully prepared cutout can work for a new background.
Keep the product and background on separate layers where the editor supports it. Match perspective and lighting around the item. Inspect the cutout so thin parts are not lost and old background fragments are not left behind.
For this route, see changing the background while keeping the product. It is more specific than asking a model to recreate the whole photograph “more professionally.”
Use a local edit for a defined correction
In Dreamina, give the smallest useful change request and an accurate reference. Name the region and the features that should remain visually consistent. Where the current editing mode provides a selection or mask, use it to express the intended area.
A prompt might say:
Change only the front pocket to warm terracotta. Keep the handles navy. Preserve the pocket outline, stitching, brass snap, cream body, bag proportions and overall lighting. Add no new seams, hardware or lettering.
Review adjacent areas as well as the target. If a repair changes the handle attachment or creates a new seam, it has introduced another problem. Avoid repeatedly editing the latest unapproved result; return to the approved source when drift accumulates.
Use reconstruction for a creative scene you can verify
A reference-led generated scene can be valuable for a campaign concept. Check it against the real product before using it as final artwork. If the scene works but the item is wrong, keep the creative direction and rebuild around the original product photograph.
Do not treat an invented rear view as documentary evidence of a side you never supplied. Photograph that view if shoppers need to inspect it.
What can a dedicated product-repair tool do?
Start with Dreamina when the task needs creative editing and a reviewable visual result. For a dedicated correction route, Photoroom offers Product Fixer, alongside its guide to inaccurate AI product images.
For targeted repair, compare the generated image with the product reference and identify the detail to correct. Photoroom’s Product Fixer is available on Pro, Max, Ultra and Enterprise, with its own credit conditions; it offers another route for product-focused revisions.
Recheck the corrected detail and its surroundings against the source. If the correction still invents construction or rewrites a label, retaining the original photographed product may be the better route.
Check the export in the place it will be used
Inspect the downloaded file at full size for detail, then at the intended display size for usability. The two checks answer different questions.
At full size, compare labels, seams, edges, surface patterns and components. At display size, check whether the item remains identifiable, important information is readable and the crop preserves the useful features.
For a white listing image, verify the actual background rather than judging it by appearance alone. Our separate white-studio example looked white but measured RGB 252 at five background points, which failed the strict 255 requirement. The white-background guide explains that distinction without treating a color check as complete product acceptance.
For video, examine the product throughout the clip. Watch the full export and inspect relevant frames where the item turns or interacts with something.
When should you stop repairing and take another photograph?
Take another photograph when the required detail is missing from the source, when the new angle matters to a buying decision, or when repeated edits keep producing a plausible substitute for the real construction.
A clear new source image can be more reliable than a complex prompt. This is particularly true for tiny lettering, transparent packaging, reflective surfaces and products with intricate patterns.
Keep the accurate listing asset separate from exploratory campaign imagery. Both can be useful, but they serve different purposes. Our product-photo set workflow shows how to plan those roles before generating a folder of interchangeable images.
FAQ
Can AI change details even if I only ask for a background edit?
Yes, depending on the workflow. A reference-led reconstruction may redraw parts of the product. A retained original layer gives more direct control, but edges, reflections and compositing still need review.
How can I keep the rest of the product unchanged?
Keep the original product on its own layer and limit the correction with a selection or mask. Use a reference-led generation when you want a creative variation; return to a source-product composite when the listing needs the original details.
Is a more realistic image necessarily a more accurate product photo?
No. Realistic lighting and texture can make an incorrect label or invented component look convincing. Compare the output with the actual item, not only with a general idea of good photography.
What is the safest starting point for a real listing?
An accurate photograph of the actual product and variant. Preserve that source, keep important product information under control and use AI to improve the presentation around it where appropriate.