Bulk Listing Images for a Catalogue

Half of this workflow is genuinely batched and half is not, and pretending otherwise would waste your afternoon. Research is bulk: import a Seller Central export, up to 300 ASINs land in the composer, and each one becomes its own background job at 0.1 credit. Image generation is per product — you open a SKU and run its set. If you want one instruction to turn forty ASINs into forty finished sets, that is the public API, and it needs someone who can write a loop.

What is batched, and what is not

The intake half, which is the good half

Export your listing report from Seller Central, drop the file into the import control, and every ASIN in it is pulled out and stacked in the composer. The parser reads /dp/, /gp/product/ and /gp/aw/d/ links as well as bare ten-character codes, and when real ASINs are present it ignores the SKU column rather than submitting your internal codes as products. Duplicates collapse. You edit the list before anything is charged.

Submitting queues one research job per ASIN at 0.1 credit each — forty SKUs cost 4 credits, about €1.33 at pack pricing. Each job scrapes the live listing, its reviews and its category best sellers, then writes a brief with pain points, USPs and six to seven image scripts. Behind the scraping proxy four of those run at once; the queue is durable, so you can close the tab, and a job that fails refunds itself automatically.

This is the step that makes catalogue work feasible at all. Research is priced deliberately low precisely so that researching everything and generating selectively is the rational order of operations.

The generation half, honestly

There is no button that says generate images for all selected products. You open a product, pick a model and a style mode, and run its set; the app allows two image jobs at once, so you can start the next product while the first renders. For a dozen SKUs that is fine. For eighty it is an afternoon of clicking, and you should use the API instead.

One call to the infographics endpoint takes an ASIN plus a country, or a listing URL, and chains the whole thing: research, then six images, then a status URL with signed links to each file. Three runs can be in flight per account, so a simple script that keeps three in the air processes a catalogue overnight without supervision. A listing already researched skips straight to the images unless you ask for a fresh brief, which matters when you re-run a set after changing the brand kit.

Variation families

The research pass scrapes the colour twister along with everything else, so the child colours of a listing are known. What it does with them is narrower than the phrase family handling suggests: it does not research each child ASIN and produce a set per child. It repaints. Select finished images from the parent's set, pick colours from the twister or type your own, and each pairing is reissued with the product in that colour and every other pixel held still — same layout, same text, same background, same shadows.

Colour accuracy comes from a reference photograph rather than a name: your own upload if you give one, otherwise the variant ASIN's main image, otherwise the swatch thumbnail. Up to 12 images by 8 colours in a run, each repaint priced as one image on the default model, landing as its own labelled gallery set. For a family that differs only in colour this is the cheapest correct answer. For a family that differs in size, capacity or contents, it is the wrong tool, and those children need their own sets.

What forty SKUs actually cost

Research for forty ASINs is 4 credits. A six-image set is 6 credits on the one-credit models, 12 on the default Nano Banana Pro, 18 on the most expensive. Forty sets therefore run 244 credits at the low end and 484 at the default, which is roughly €81 to €161 at pack pricing, or the Growth plan at €199.99 a month for 875 credits if this is a recurring job rather than a one-off. Compare that against the freelance floor of $10 to $15 an image: forty six-image sets is 240 images, which is $2,400 at the very bottom of the market, before briefing time and revision rounds.

The comparison is not free of caveats. A freelancer delivers art direction; this delivers a consistent system quickly. At five SKUs the price gap narrows and the quality argument favours a good designer. At forty it does not.

How to run a catalogue

Set the brand kit first, because it applies at generation time and cannot be applied retroactively. Import the export, research everything, then read the briefs before spending image credits — the brief is where you find out that half your catalogue shares one pain point and should share one comparison image. Generate one product completely, look at it at thumbnail size, adjust the kit or the style mode, and only then work through the rest. If the rest is more than fifty SKUs, or you will do this again next quarter, spend an hour on the API instead of a week in the interface.

Common questions

Can Graflio generate images for many products at once?

Not from the interface. Research is submitted in bulk and runs several jobs in parallel, but image sets are started per product, with two jobs running concurrently. The public API runs a whole ASIN-to-images job in one call, three at a time per account, which is how catalogues are processed unattended.

How many ASINs can I import at once?

A CSV or TXT file up to 5 MB, from which up to 300 ASINs are parsed into the composer. They are submitted in batches of 20, and each account holds 200 saved products.

What does bulk research cost?

0.1 credit per ASIN — 40 ASINs for 4 credits, roughly €1.33 at pack pricing. Images are charged separately, at 1 to 3 credits each depending on the model. Any job that fails is refunded automatically.

Does it handle variation families?

Partly. The colour twister is scraped, and finished images can be repainted into the family's other colours, up to 12 images by 8 colours per run, with everything except the product colour held identical. It does not research each child ASIN separately, so families that vary by size or contents need their own sets.

Will a catalogue look machine-stamped?

That is the real risk, and the defence is the style mode. Each mode rotates a different treatment per image — flood, macro, split, lifestyle, exploded view, dimensions, cutaway — so a set reads as a sequence, while two SKUs in the same mode still look like siblings. The copy differs because each brief is built from that listing's own reviews and competitors.