The prompt structure
| Part | What to write | Example |
|---|---|---|
| Product | Exact item, color, material, key details | a sand colored linen midi dress with wooden buttons |
| Setting | Background or scene | on a seamless light grey studio background |
| Light | Type and direction of light | soft diffused daylight from the left, gentle shadow |
| Camera | Angle, distance, framing | straight on, full length, product centered, square frame |
| Style | The look of the final image | clean ecommerce catalog photo, realistic, sharp focus |
| Constraints | What must not change | keep the exact color, buttons and length from the reference photo |
Write the parts in this order. Most image models give more weight to what comes first, so the product belongs at the start. The examples in the table are illustrations we wrote, not output from a specific tool.
Always start from a real product photo
The biggest mistake with AI product photography prompts is describing a product instead of showing it. A text prompt for a linen dress produces a linen dress, but not yours: the buttons, the neckline and the exact shade will be the AI's idea of a linen dress. For a store, that image is useless, because it shows a product you do not sell.
So use tools that accept a reference image, and attach a clear photo of the real product. Shoot it in daylight, on a plain background, with the whole product in frame and no filters. Then the prompt describes what to do with the product, not what the product is. Many general AI assistants accept an uploaded image with a prompt; check what yours supports.
Example prompts by image type
These are example prompts written for this article, to show the structure. Adapt the product description and keep the constraints.
Clean main image: "Using the attached photo, create an ecommerce product image of this black leather crossbody bag on a pure white background, soft even studio light, subtle shadow under the bag, straight on, product filling most of the square frame. Keep the exact shape, stitching, strap and metal hardware."
Lifestyle scene: "Place the attached ceramic coffee mug on a light oak table next to a window, soft morning light from the right, shallow depth of field, a linen napkin in the background out of focus. Keep the mug's glaze color and logo unchanged."
Flat lay: "Create a top down flat lay of the attached knitted sweater on a light beige linen surface, sleeves folded neatly, soft shadowless light, square frame with some empty space around the garment. Keep the knit pattern and color exactly as in the photo."
On-model image: "Show the attached denim jacket worn by an adult woman standing in a minimal light grey studio, full length, front view, natural relaxed pose, soft studio light. Keep the wash, the buttons, the pocket placement and the length of the jacket exactly as in the photo."
Words that help and words that hurt
Some words reliably push results in a useful direction:
- For realism: realistic, natural light, true to life color, sharp focus, ecommerce photo.
- For framing: full length, three quarter view, close up of the fabric, centered, square frame.
- For light: soft diffused light, window light, even studio light, gentle shadow.
- For control: keep, unchanged, exactly as in the reference photo.
Other words cause trouble. Cinematic, dramatic, artistic and moody tend to change colors and add heavy shadows, which hides the product. Perfect, flawless and beautiful add little and can smooth fabric texture. Brand names of famous photographers or magazines are best avoided: they shift the style unpredictably and raise questions you do not need.
Length matters less than order and clarity. A prompt of two or three precise sentences usually beats a long list of adjectives, because every extra word is one more thing the model may misread.
Fixing the most common problems
When the result is wrong, change one thing at a time and generate again.
- Wrong color: add the color twice, once in the product description and once in the constraints, and mention the reference photo. Check the input photo too: if it has a yellow cast, the output will follow it.
- Changed print or logo: say "keep the print and logo exactly as in the reference photo, do not redraw them". If it still changes, crop a close up of the print and attach it as a second reference.
- Product too small in the frame: say "the product fills most of the frame" and name the framing, such as full length or close up.
- Extra objects: add "no other products, no props" or "only the background".
- Inconsistent catalog: save the exact prompt and reuse it word for word for every product in a collection, changing only the product description.
Keep a simple text file with the prompts that worked. A prompt library is what turns one good image into a consistent catalog.
When a form beats a prompt
Prompts give general AI tools a lot of freedom, and freedom is the problem for a catalog. Every word changes the result a little, so two products prompted on different days rarely match. That is fine for a social post and hard for a category page.
Dedicated product photo tools replace most of the prompt with choices. In Wizia AI Studio you upload the product photo, pick its category, choose an AI model from the library or switch the model off, pick a scene and optionally a pose, and set the quality. The same choices give the same look for the next product. A 2K image costs 1 credit and a 4K image costs 2 credits.
On Pro+ and higher plans you can also create a custom model and custom scenes by describing them in a short text, written in English, and saving them to your library. That is the one place where prompt writing skills still help, and you only write the description once.
General AI or a fashion tool?
General AI assistants are flexible and often already part of a subscription you have. They work well for ideas, social images and products that do not need to be worn. For clothing on models, the risk is that the garment changes between images, and checking every detail takes time.
A fashion tool is built for one job: keep your garment accurate and show it worn, in the same style across the catalog. If you sell clothing on Shopify, WooCommerce or CloudCart and publish new products every week, that consistency usually matters more than creative freedom. We compare both routes in Wizia vs ChatGPT and Gemini, and our guide on creating product photos with AI covers the full workflow.
A checklist before you publish
- The product in the image matches the real product: color, print, shape, details.
- The framing is the same as the rest of the collection.
- No extra objects, text or watermarks you did not ask for.
- Hands, edges and straps look natural at full zoom.
- The prompt is saved for the next product.
The Free plan of Wizia includes 20 one-time credits if you want to compare a form based tool with your prompts on the same product; its images carry a watermark. Starter costs €20 a month with 150 credits. See the pricing page and the AI tools hub.