Batch product photos for a 50-SKU store in one afternoon with AI

The set: four images per SKU, no more, no fewer

An afternoon batch cannot carry a ten-image set, and it does not need to. Four images answer the four questions every shopper asks, in the order they ask them. The hero on white answers what exactly am I buying. The lifestyle image answers would this fit my life. The close-up answers is it well made. The scale image answers how big is it really. Everything else, unboxing, variants, flat lays and extra lifestyle scenes, is a second batch for best sellers, not a reason to delay fifty listings. A store with fifty complete four-image listings sells; a store with ten perfect ten-image listings and forty empty ones does not.

Write one product master per SKU before you generate anything, and use it unchanged in every prompt and scene. The master is one sentence: product, material, colour, size, and one distinguishing detail, for example a matte black stainless steel bottle, 750 millilitres, screw cap, small printed logo on the front. If the master changes between images, the product changes between images, and the shopper notices even if they cannot name why. Our ten Shopify scene prompts use the same discipline for a full set; this batch uses the first four of those jobs, in the same order, because order is what keeps a catalogue consistent when you are moving fast.

The afternoon schedule

Grouping is the step that makes the schedule real. Do not run SKUs in catalogue order if that order mixes a small black bottle, a large beige tote and a reflective steel jug, because every switch forces a new lighting and framing decision. Group by size first, then by material within size: small matte items together, large fabric items together, reflective items together, and run each group straight through. The same logic drives professional batch shooting, where locking exposure and white balance and swapping items without re-metering cuts per-SKU time from minutes to seconds. With AI generation the saving is smaller per image but larger in consistency, because a locked prompt and a locked group produce a grid that looks like one store.

Naming, export and the cost maths

Name files by SKU from the first save, never by a descriptive title you will have to rename later. A workable pattern is SKU, angle and version, for example a file named for the SKU followed by hero, lifestyle, detail or scale, and a version number for regenerations. Keep the version number, because when a hero fails QC and is regenerated, you need to know which file is the keeper without opening both. Store one folder per SKU, or one folder per angle with SKU filenames, but pick one system and do not mix them mid-batch. A file attached to the wrong product is the most expensive mistake in a batch, because it is invisible until a buyer complains.

Export the hero at 2048 pixels on the longest side, lifestyle and scale images at the same resolution, and compress for the web only after the final edit, so compression never hides a flaw you would have fixed. Square, 1 to 1, is the safe default for heroes, because it crops cleanly in collections and marketplaces. Portrait, 4 to 5, suits lifestyle images on mobile feeds. Check the five spot-check listings on your own phone at the end, not on a desktop preview, because the phone is where the order happens: hero readable as a thumbnail, lifestyle image desirable rather than merely descriptive, scale image removing the size doubt without reading the description.

The generation maths is refreshingly small, which is why the afternoon is spent on decisions, not on budget. Two hundred images at about 0.8 cents each cost about 1.60 dollars. Budget double for regenerations, about 3.20 dollars, and you will still finish under five dollars for the whole catalogue. Compare that with third-party guides that put a traditional studio image at 25 to 100 dollars each, which would price the same 200 images at 5,000 dollars and more, before studio time. The AI batch does not beat a studio on every image, and a hero that drifts in colour or shape should still be photographed for real, but for a 50-SKU launch the comparison is not close. Spend the saving on the QC pass, because a cheap batch published without checking is the most expensive batch of all.

Two-pass QC, and when to stop a SKU

Pass one is fast and covers all 200 images at thumbnail size. It catches the failures that cost sales at a glance: wrong colour, wrong shape, product cropped, product lost in the background, halo from a bad cutout, text invented on a label. Any image that fails pass one is regenerated once, with the same master and a tighter prompt, and if it still fails, the SKU is marked for a real photograph or a background swap, not given a third generation on the same prompt. Pass two is slow and covers only the 50 heroes at 100 percent. It catches grain mismatches, warped logos, broken seams and soft focus, the flaws that survive a thumbnail and ruin a zoom.

Stop rules keep the afternoon on schedule. If three images in one group fail the same way, stop the group and fix the prompt or the source photo, because the next twenty will fail the same way. If one SKU fails three times, park it and finish the other forty-nine, because a catalogue that is 98 percent live beats a catalogue that is 100 percent perfect at midnight. If a product's colour drifts in every generation, photograph it for real: AI is a batch tool, not a colour-correction service, and a buyer who receives a different colour will return it whatever the photo cost. The remove-background guide covers the cutout step, and the small-shop photo guide shows how the same four-image set scales to a second batch for best sellers.