Batch Generation: How to Build a 100-Item Product Catalog Without Losing Your Mind

One product listing is a nice evening project. A hundred of them is a different kind of job, the kind that wears people down. The difference isn't about skill, it's about how you approach the work. Generating images one at a time falls apart once volume shows up, while batch generation turns the same work into an assembly line. Here's how I run the process when I've got dozens or hundreds of SKUs to get through.

Fix Your Source Files Before You Touch Generation

The biggest mistake at scale is jumping straight to generation. Sort out your inputs first: product photos shot (or at least prepped) the same way, one consistent spec sheet, file names that make sense at a glance. An hour spent cleaning up your source material saves you a full day of redoing work, because a model fed messy inputs gives you messy outputs, and you won't notice until you lay the whole batch out side by side.

One Template, One Series

Next comes the template: a fixed prompt where only the product itself changes, while lighting, background, angle, and style stay locked in place. That speeds things up and gives your whole catalog a matching look, which matters if you're selling on Amazon or Etsy. I got into why style still drifts even when you think you've locked it down in my piece on seed values and reproducible generation.

Batch Generation: Run the Whole Series in One Sitting

Models drift more between sessions than within a single one. So a hundred listings are better handled in two or three long sessions than five a night for two weeks straight. Working this way also keeps the whole job in your head, which means you catch problems faster.

Review Is Its Own Job

At scale, going through the output takes longer than generating it, and it deserves its own slot on the schedule. I generate with some margin built in, then go through the whole batch in one pass and cut anything that doesn't work instead of trying to patch every bad frame. Regenerating is cheaper than fixing.

Where the Pipeline Breaks Down

Let me be honest about the rough edges. Products with text printed on the packaging still need a manual check; running them through on autopilot is risky. Trickier categories like jewelry or shoes produce more rejects on a production run than plain flat-lay items do. And there's eye fatigue: after a hundred frames you stop noticing artifacts, so review works better with a fresh head rather than right after a long generation session.

I won't pretend I've found the perfect balance between speed and quality here. Big catalogs always force some kind of tradeoff. You can build a series of listing images straight from product links at Flami, but still budget separate time for prepping your source files and for reviewing what comes out.

FAQ

Where do I start when I'm facing a big catalog?

With the source files, not the generation itself: photos prepped the same way, one consistent spec sheet, file names that actually make sense. An hour on the input side saves you a full day of rework, because on messy inputs the flaws only show up once you've already finished the whole batch.

Why run a series in one sitting instead of spreading it out?

Models drift more in style and color between sessions than they do within one. Running it in one sitting also keeps the context in your head, so you spot systemic errors faster than you would working five listings a day across different sessions.

How much time does review actually take?

At scale, more than the generation itself, which is why it needs its own slot on the schedule. It's more practical to generate with margin and go through the batch in one pass, cutting the rejects, instead of trying to fix every weak frame individually.

Which products don't handle batch runs well?

Anything with text on the packaging, plus trickier categories like jewelry or shoes: the reject rate there runs noticeably higher than on plain flat-lay product shots. It's worth pulling those out of the main pipeline and checking them by hand.

Sources

1. Flami: seed values and reproducible generation
2. Flami: a month-long content plan for your store