product image consistency checklist: measure the catalog rather than matching impressions
A product image consistency checklist should describe things a reviewer can measure. Asking for the same feel across a catalog leaves too much room for interpretation. Define the product’s occupied region, camera family, background treatment, and export rules before comparing generated or edited candidates.
Genace can supply image candidates, but a shared model or prompt does not guarantee consistent framing across SKUs. Alignment, crop normalization, color review, and final catalog assembly are application-side and editorial responsibilities.
Define a reference tile
Choose one approved catalog image and document why it works. Record the canvas dimensions, subject bounding box, baseline, top margin, and intentional shadow style. Keep the example alongside a simple diagram rather than relying only on prose.
Different product families may need different reference tiles. A tall bottle and a wide tray should not necessarily occupy the same bounding box. Create a small number of documented families, then map every SKU to one family so variation is controlled rather than accidental.
Use a review matrix
| Dimension | Example internal rule | What to inspect |
|---|---|---|
| Framing | Product fits the family guide | No accidental clipping |
| Baseline | Contact point aligns to guide | Product does not jump in a grid |
| Background | Approved neutral treatment | No unexpected color cast |
| Shadow | Consistent direction and softness | No mixed lighting story |
| Label | Readable at detail size | No altered product wording |
These are example house rules, not Genace settings. Choose values based on your actual storefront and test them against representative products. Store the rule version with each approved export.
Separate permissible variation
Some differences communicate real product facts. A larger package should not always be scaled down until it appears identical to a smaller one. Decide whether a tile normalizes composition or communicates relative size, and label that decision in the catalog specification.
Review color using approved references. Do not force two genuinely different finishes into the same color appearance merely to make the grid uniform. Consistency should improve comparison while preserving meaningful differences.
Use a contact sheet containing a mix of old approved images and new candidates. Check rows and columns for visual jumps, then inspect any flagged image individually. A contact sheet reveals alignment problems but cannot verify small structural or textual details.
Repair through deterministic editing
Where appropriate, adjust canvas placement, crop, or background in an editor. Keep the real product layer unchanged. If a generated candidate has a distorted silhouette, reject it instead of treating deformation as an alignment correction.
Record the change as a derivative with its source file and recipe. This helps distinguish an approved source from a delivery crop that later becomes outdated when the storefront layout changes.
Acceptance checklist
Review a full collection grid at its actual display width. Confirm family mapping, margins, baseline, color references, label accuracy, and crop behavior. Verify that the asset database points to the reviewed derivative for each SKU and that exceptions have named approval rather than unexplained inconsistency.
Related guides cover background production, variant evaluation, and responsive derivatives.
Continue with creative workflows
Where these facts come from
- codebase: src/app/api/v1/images/generations/route.ts
- codebase: src/ai/providers/nano-banana.ts
- Application-owned workflow recommendations; human review and post-processing are outside the generation gateway
- Editorial keyword hypothesis; search volume not measured