Generate product images from a shared visual brief
Use approved product references and photography guidelines to prepare consistent images for a store or campaign.
How this works
Use approved product references and photography guidelines to prepare consistent images for a store or campaign. Product imagery must preserve the actual product even when the setting is generated. Shape, color, labels, included components, and scale can affect a buyer's expectations, so visual consistency is also a factual accuracy problem.
Use approved product references as the fixed anchor and vary only the permitted scene attributes. Generate a small set of candidates, compare them with the reference, and provide a designer with clear acceptance and rejection notes. Keep image production separate from listing publication.
Before you start
Accurate product references
Provide multiple useful views and known specifications.
Photography rules
Define background, lighting, angle, crop, and acceptable props.
Placement requirements
Specify dimensions, format, and whether the image is illustrative or a product listing.
Prepare the inputs
- product_assets
- Required
Approved images
Use the correct SKU and version.
- product_constraints
- Required
Fixed features
List colors, geometry, labels, and included items.
- scene_brief
- Required
Setting and composition
Separate permitted changes from fixed product details.
- brand_style
- Required
Visual rules
Use current guidance.
- export_requirements
- Required
Dimensions and formats
Identify crops needed for each placement.
Implementation, step by step
Inspect the product anchor
- Review reference views and list distinctive details that must survive generation.
- If a side or component is not shown, avoid inventing it.
Checkpoint: A product-fidelity checklist.
Specify the scene
- Define camera angle, lighting, background, and props within the brief.
- Keep scale and physical relationships plausible without adding unsupported accessories.
Checkpoint: A controlled generation brief.
Generate candidates
- Use available image tools with the approved references.
- Label each candidate by scene and version for later review.
Checkpoint: A manageable candidate set.
Check product fidelity
- Compare shape, color, text, components, and material details with the original.
- Reject subtle changes that could mislead a buyer, even if the image looks attractive.
Checkpoint: A fidelity-reviewed shortlist.
Refine layout and crops
- Prepare required placements with safe margins and readable product details.
- Use controlled overlays for exact text rather than accepting distorted generated labels.
Checkpoint: Placement-specific drafts.
Prepare the asset handoff
- Export actual files and record approved and rejected variants.
- Provide the listing or campaign owner with the fidelity checklist and remaining limitations.
Checkpoint: A reviewable image package.
Decisions along the way
A generated image changes a feature
Regenerate or use a real cutout rather than hide the discrepancy.
Reference coverage is insufficient
Request the missing view or choose a scene that avoids it.
A prop could imply inclusion
Remove it or clarify the visual according to the placement brief.
What you will produce
product-image-drafts
Image filesProvides consistent candidates for review.
Include: sku · scene_id · version · dimensions · status
product-image-review.csv
CSVDocuments why an image is usable.
Include: candidate_id · feature_check · issue · decision
Check the result
- Compare the selected image with the exact SKU references.
- Inspect labels and small components at full resolution.
- Check scale and included-item implications.
- Review every crop rather than assuming one composition fits all placements.
When something goes wrong
Generation repeatedly changes geometry
Use a reference-preserving composition or manual product cutout.
Brand style overwhelms product clarity
Simplify the scene and restore product visibility.
Export resolution is inadequate
Regenerate or export at the required size without claiming detail that is absent.
A worked example
SYNTHETIC EXAMPLE · NOT A REPORTED BUSINESS RESULT
Request
Create a consistent store image for a blue bottle.
Inputs
Synthetic example. Synthetic reference shows a plain cap and a 500 ml label; generated candidate adds a pump and changes the label to 750 ml.
Expected result
Synthetic expected result. Reject that candidate. The approved version preserves the cap and label while changing only the background and lighting.
Run a small pilot
Start small
A few SKUs with distinct shapes and label details.
Record the baseline
Record candidate rejection reasons and designer correction time.
Agree success criteria
Agree product-fidelity and placement-quality criteria before batch production.
Stop or revise
Stop listing use when an image changes a material product attribute.
Keep the scope clear
- Do not invent features, included accessories, or certifications in imagery.
- Generating an image does not authorize publishing it to the store.
Check before using
Check product shape, details, colors, and any visual claims.
Business activities
- Create and publish content
- Manage product catalogs and storefronts
FROM A RECIPE TO YOUR WORKFLOW
Bring the company context together.
This workflow needs reliable inputs before it needs a better prompt. Moat8 Vault brings working information into a readable archive for the AI tools you use.
For this recipe: Product reference images · Brand palette · Photography guidelines
Start with one recurring task, agree which sources to use, and review the first result with your team.
Discuss this workflow