A Catalog-Grid Export Workflow

A product catalog grid is one of the most technically demanding asset types in e-commerce design. Every image in the grid must read as part of a unified visual system — identical background tone, matching shadow depth, consistent product scale relative to the canvas, and uniform whitespace around each product. When even two or three images in a 48-product grid have a slightly warmer background or a shadow that falls at a different angle, the grid reads as assembled from multiple photo shoots rather than as a coherent product family. That inconsistency signals low production quality and undercuts the premium positioning most catalogs are trying to establish.

The challenge is compounded by the typical origin of catalog source images. Product photographs come from multiple photo shoots, different photographers, different studio setups, and sometimes different countries. The hero shot of a sofa might be lit from the left with warm tungsten; the hero shot of a lamp might be lit from the right with cool LED. Before any layout can happen, every image must be normalized to a shared visual specification. This normalization step — background replacement, shadow generation, color temperature correction, scale normalization — is where most catalog production time is lost.

A Floniks catalog-grid export workflow automates the entire normalization and layout pipeline. Source images enter the workflow in batch; each passes through a normalization chain that enforces the shared visual specification; normalized images are assembled into a configurable grid layout with consistent gutters and margins; and the completed grid is exported in both print-resolution and web-optimized formats. What previously required a design team working for days takes a matter of hours with the workflow running.

The workflow begins with a Batch Image Ingest node configured to accept a folder of product images. The node reads each image file and emits one execution token per image, carrying the image data and the filename. Downstream normalization nodes receive the tokens in parallel batches, processing multiple images simultaneously rather than sequentially.

The first normalization step is background replacement. Each product image passes through a Background Removal node that isolates the product on a transparent alpha, followed by a Background Synthesis node that generates the approved catalog background. The background prompt should be locked in the node configuration: "pure white studio background with soft directional shadow falling at 45 degrees below and to the right, subtle ground contact shadow, professional product photography, no texture, no gradient except the natural shadow." This prompt never changes between runs, so every product in the catalog receives exactly the same background treatment.

Scale normalization follows. A Scale Fit node receives the product image on its transparent background and scales it so the product occupies between 70% and 80% of the canvas height, centered horizontally, and with the product base resting at a consistent vertical position (typically 80% from the top of the canvas, leaving 20% for the shadow to land below the product). This rule ensures that a small coffee mug and a large armchair appear at visually equivalent scales within their individual tiles — neither looks disproportionately large or small in the grid. The output of the Scale Fit node is a 1:1 square tile at the target catalog resolution: 2000x2000 pixels for print, with web derivatives generated later.

After scale normalization, each tile passes through a Shadow Refinement node. The Background Synthesis node generates a base shadow, but the shadow quality benefits from a dedicated refinement pass that sharpens the contact shadow beneath the product and softens the cast shadow at its far edge. The shadow prompt: "soft drop shadow, hard contact edge beneath the product base, shadow opacity 40%, shadow spread 120 pixels, shadow angle 315 degrees, no color cast in shadow, pure neutral gray." These values should be stored in a shared Shadow Config node so every product in the batch receives identical shadow treatment.

Color temperature correction is the next pass. Products photographed under different studio lighting conditions will have slightly warm or cool casts even after background replacement, because the product surface itself carries the color temperature of the original light source. A Color Normalize node applies a neutral color correction to each product image: "correct color temperature to D65 daylight neutral, remove any warm or cool cast from the product surface, preserve accurate product color rendering, do not oversaturate." This pass brings all products to a common chromatic baseline without altering the actual product colors — a navy jacket remains navy, a cream sofa remains cream, but incidental lighting casts are removed.

Run a quality check pass on a representative sample before committing to a full-batch run. Select five to ten products that cover the extremes of the catalog — the largest and smallest products, the warmest and coolest original photographs, the most complex and simplest shapes. Review their normalized tiles side by side. If any normalization step is producing inconsistent results, adjust the relevant node parameters before processing the full batch. This sample-first approach catches systematic errors early and avoids wasting processing credits on a full run that requires rework.

Once all product tiles are normalized, they route to a Grid Assembly node. This node accepts a list of normalized tile images and a layout configuration — number of columns, gutter width in pixels, outer margin in pixels, and the page canvas dimensions — and assembles the tiles into the catalog grid. For a print catalog page at A4 size (2480x3508 pixels at 300 DPI), a four-column grid with 40-pixel gutters and 80-pixel outer margins produces a professional layout. For a web catalog page, a three-column grid at 1440 pixels wide with 20-pixel gutters and 40-pixel outer margins is appropriate.

The Grid Assembly node places tiles in reading order (left to right, top to bottom) unless a custom ordering sequence is provided. If the catalog has a specific product ordering based on SKU priority, price, or category, provide the ordering as a JSON array of product IDs in the Ingest node configuration. The assembly node reads the ordering and places tiles accordingly, regardless of the alphabetical or numeric order of the original filenames.

For catalogs that include category headers, section dividers, or promotional callout tiles within the grid, add a Layout Interrupt node between the Ingest and Assembly nodes. The Layout Interrupt node accepts a manifest file that specifies the grid position and content of any non-product tiles — a category header spanning the full width at position 5, a "New Arrivals" badge overlay on specific product tiles, a promotional callout tile at position 13 in the sequence. The Assembly node reads these layout interrupts and incorporates them into the grid without displacing the surrounding product tiles.

The completed grid canvas passes to a dual-export chain. The print export branch produces the grid at full 300 DPI resolution in PDF/X-1a format, which is the standard prepress format for commercial print. The node applies CMYK color space conversion using the ISO Coated v2 profile (or the profile specified by the printer). Page bleed and crop marks are added at the standard 3mm specification if the print template is configured for full-bleed layout. The print PDF filename follows the convention: "catalog-grid-[section]-[version]-print.pdf."

The web export branch produces the same grid canvas at 72 DPI in JPEG format at quality 88, which balances file size against visual fidelity for on-screen viewing. A WebP variant is also produced at the same resolution for modern browser contexts. The web filenames follow the convention: "catalog-grid-[section]-[version]-web.jpg" and "catalog-grid-[section]-[version]-web.webp."

Individual normalized tile images are exported alongside the assembled grids because they are independently useful for product detail pages, ad creatives, and marketplace uploads. The tile export node names each file with the product SKU: "[SKU]-normalized-tile.jpg." For catalog teams that also need tiles at specific platform dimensions — 800x800 for Amazon, 1000x1000 for Shopify, 600x600 for a third-party marketplace — add a Platform Resize node downstream of the tile export that produces each platform variant with the appropriate naming suffix. Save the complete workflow as a named template: "CatalogGrid-[Season]-v1." Refresh runs for the next season require only swapping the batch ingest folder and updating the layout configuration if the column count or page dimensions change.