Product Imagery at Catalogue Scale

Generating one good product image is a creative problem. Generating ten thousand is a logistics problem: getting the products in, keeping track of which ones are done, keeping brands apart, and keeping a shared team working from the same library.

Getting products in

Three routes, no integration work:

• Feed URL: paste a link to an XML or CSV product feed
• File upload: drop an XML or CSV file directly, up to 20 MB
• Google Merchant Center: connect over OAuth and pull the catalogue

XML support covers RSS, Atom and the Google Merchant g: namespace. Columns are auto-detected where they can be, and anything unusual can be mapped by hand during import. SKU and title are the only required fields.

Multi-market catalogues

Feeds in 23 languages are detected automatically, from g:content_language, g:target_country, a <language> tag, or URL patterns like feed-sv.xml and ?lang=sv. The list leans European and Nordic: Danish, Finnish, Norwegian, Swedish, Estonian, Latvian, Lithuanian, alongside German, French, Spanish, Italian, Dutch, Polish and the rest.

Group feeds into a catalogue per market and the same SKU is connected across every language feed, so one product carries all its localised variants rather than appearing as five unrelated rows.

Keeping thousands of images findable

Everything generated lands in a shared team gallery: searchable by label, description or the prompt that produced it, organised into folders, and reusable as input for the next generation. At catalogue scale the library is the product. An image you cannot find again is an image you will regenerate and pay for twice.

Agencies and multiple brands

Each brand gets its own company workspace. Products, AI models, brand colours and generated images stay cleanly separated per client, while one login moves between them. A model built for one brand does not leak into another's library, and neither do the images.

Consistency across a long tail

The two things that make a large catalogue look coherent are the same setup applied everywhere and a house style that is easy to reuse:

• Blueprints bake a full setup (prompt, composition, aspect ratio, quality tier) into one click
• Team blueprints let you publish your own configured workspace so everyone starts from the same house style rather than approximating it
• Brand Kit pins colours to exact hex values, autocompletable inside prompts

Cost control

Credits are consumed per generation, priced by model and quality tier. Drafts cost a few credits, standard images roughly 11 to 16, and premium 4K renders up to about 45, so a 2,750-credit plan is good for roughly 170 to 250 standard images.

At catalogue scale the practical pattern is to iterate at draft quality until the look is right, then regenerate the approved setup at standard or high. Getting the blueprint right once and applying it broadly costs far less than exploring on every SKU.

Disclosure at scale

Every export carries a signed C2PA manifest and IPTC AI-generated labelling by default, which is the part that matters when you are publishing thousands of images into channels that increasingly check for it. There is nothing to remember and nothing to apply per image.