Craft fair season has a way of exposing every weak spot in your photo setup. You spend three weeks making forty candles, then realise the only pictures you have were taken on a kitchen table at 11pm with a phone balanced on a mug. Your stall banner looks fine. Your price cards look like afterthoughts. And by the time you get home, you have 200 shots and none of them match each other.
This is where a generative tool earns its keep. If you have been putting off learning one of the many AI image generators because the marketing makes it sound like magic, here is the honest version: it is a fast, cheap way to produce consistent product images at volume. Not perfect reproductions of your actual objects — that distinction matters and we will get to it — but the supporting visuals that make a stall look professional.
Start with the images you actually need
Before opening any tool, list the assets by size and purpose. A craft stall needs five kinds of image and they are all different jobs:
First, a wide lifestyle shot for the banner behind your table. Second, square product shots for price cards. Third, a flat-lay hero image for the online listing. Fourth, in-context shots showing the item being used or worn. Fifth, small detail crops for the "how it's made" board.
Most people try to generate one beautiful image and stretch it across all five. That never works because the aspect ratios fight each other. Decide the sizes first — banner, square card, portrait listing, wide context — and generate to those shapes from the start.
Keep a written style recipe
The single biggest cause of an inconsistent stall is that makers rewrite their prompt from scratch every time. You get one image in soft morning light and the next in harsh midday sun, and the table looks like a jumble sale.
Write your recipe down once and reuse it. A workable structure is: subject, material, surface, light direction, colour mood, camera distance, aspect ratio. Something like "hand-poured soy candle in a matte amber jar, on reclaimed oak, single soft window light from the left, warm neutral palette, medium close-up, square". Change only the candle. Keep everything else locked.
Run six variations of the locked recipe back to back and you will see how much the container and the colour story start to agree with each other. That agreement is what makes a stall look curated instead of cluttered.
Be careful with your real products
Here is the part most tutorials skip. If you are generating images of things you genuinely sell, the image has to survive the customer meeting the real object. Generated wood grain is often smoother than real wood grain. Generated knit texture is often too regular. Generated glaze colours drift.
The safe split is this: generate backgrounds, props, signage, packaging mockups, mood boards and the pattern on your price cards. Photograph your actual products. Then combine the two — cut your real photographed item onto a generated surface. You get the polish without misrepresenting the thing in the customer's hand.
Prices are a separate warning. Never let the tool render numbers. It will invent plausible-looking digits and you will only notice after printing thirty cards. Add prices in your design app, not in the generator.
Batch by collection, not by item
A craft fair is sold as a collection even when the pieces are unrelated. So generate in collections. If you make ceramics plus printed tea towels plus candles, give each line its own palette and surface, but keep the lighting direction identical across all three.
Identical lighting is the trick that ties mismatched product categories together. Same light from the same side, same shadow softness, same background family. You can vary colour freely as long as the light stays put.
Prep for print early
Screens and printers disagree about colour and about darkness. Images that look lovely on your phone often print muddy. Work in a slightly brighter, slightly lower-contrast version than looks correct on screen, then check one test print before committing the whole batch.
Also generate at the largest size you might ever need. Signage gets reprinted bigger next season and a small source file will not stretch. It is far easier to shrink than to enlarge.
A realistic time budget
Expect the first collection to take two to three hours, most of it spent deciding on your recipe. The second collection takes twenty minutes because you are only changing the subject line. That compounding effect is the real reason to learn one of these AI image generators properly rather than dabbling.
Keep a folder of your best output with the prompts written in the filenames. Next season you will not be starting from nothing at 11pm with a phone balanced on a mug.