Product photography is the highest-volume commercial use of AI image generation. Every ecommerce store, every Amazon seller, every Shopify merchant needs hundreds of product images — hero shots, lifestyle shots, detail shots, size reference, on-model shots. A prompt library for product photography is worth real money.
This is the production guide: the 6 prompt patterns for product photography, the lighting and composition rules, the per-platform requirements (Amazon, Shopify, Etsy, Instagram), the color-accuracy rules, the batch-generation workflow, and the prompt templates you can copy and adapt.
Why product photography is a different problem
Product photography is not "make a beautiful image." It is "make an image that sells the product — accurate colors, accurate materials, accurate scale, accurate details, on the right background, with the right lighting, in the right aspect ratio for the platform."
Five things make this hard:
- Color accuracy. The product in the image must match the product in real life. Shoppers return items that look different from the photo. AI models drift on color.
- Material accuracy. A "matte black ceramic mug" must look matte, black, and ceramic — not glossy, gray, or plastic.
- Scale and proportion. The product must be the right size relative to the scene. A coffee mug must look like a coffee mug, not a teacup.
- Background control. Amazon requires pure white (#FFFFFF). Etsy allows lifestyle. Instagram wants vibrant lifestyle. The model must respect the platform's background rule.
- Batch consistency. 50 SKUs need 50 images that all look like they came from the same studio. Same lighting, same angle, same color grading.
The fix is a structured prompt pattern + a per-platform check + a batch workflow. Below is all three.
The 6 prompt patterns for product photography
These are the 6 production patterns that cover ~90% of ecommerce product photography needs.
Pattern 1: Pure white background studio shot (Amazon / Shopify hero)
The single most common product photography prompt. Hero image for an Amazon listing, a Shopify product page, an Etsy thumbnail. Pure white background, no shadow (or soft shadow), centered product, straight-on angle.
The prompt structure:
A studio product photograph of [PRODUCT], centered on a pure white
background (RGB 255, 255, 255). Soft even studio lighting with a
large softbox key from camera-left at 45 degrees and a smaller fill
from camera-right. No hard shadows. The product fills 80% of the
frame. Sharp focus on the product, slight depth of field falloff
on the background. Color-accurate to the original product
(materials, finish, color). Photorealistic, 1024x1024, no watermark,
no text, no logos, no people.
Why it works: "pure white background (RGB 255, 255, 255)" gives the model an exact color target. "Color-accurate to the original product" forces the model to think about the actual product, not invent one. The lighting setup is specific (softbox key, fill, no hard shadows).
On Nano Banana 2 specifically: excellent for this. Use the official "studio product shot" pattern from the prompt guide — "High-resolution studio photograph of [product] on polished concrete, three-point softbox lighting creating soft highlights, 45-degree elevated shot focusing on [detail]."
On GPT Image 2 specifically: strong for hero shots. Specify "no watermark, no text" — the model sometimes adds fake brand marks.
On Seedream 5.0 Lite specifically: the reasoning model is good at "color-accurate" — it interprets the materials and finish from the description.
On Grok Imagine specifically: works for simple products; may be less reliable for complex materials.
The per-platform tweak:
- Amazon main image: pure white #FFFFFF, no shadow, no text, no logos, 1000x1000 minimum, 2000x2000 recommended, product fills 85% frame
- Shopify product page: pure white or light gray background, soft shadow allowed, 2048x2048 recommended
- Etsy listing: pure white OR lifestyle, 2000x2000 minimum
- Instagram shop: 1080x1080 square or 1080x1350 portrait, lifestyle backgrounds OK
Pattern 2: Lifestyle / context shot (Instagram, Pinterest, email)
The "show the product being used" prompt. The product is in a real-world context — on a table, on a person, in a kitchen, in a gym, in a car. The product is the focal point; the scene is the supporting cast.
The prompt structure:
A lifestyle product photograph of [PRODUCT] in [CONTEXT]. The
product is the focal point, in sharp focus, with a slight depth of
field falloff on the background. The scene is [describe the scene
in 2-3 sentences]. Natural lighting from [direction], color
temperature 5500K daylight, [mood: warm/inviting/energetic]. The
product is color-accurate to the original. Photorealistic, no
watermark, no text, no logos.
Examples:
A lifestyle photograph of a matte black ceramic coffee mug on a
warm wooden kitchen table, steam rising from the coffee, soft morning
light from a window camera-left, color temperature 5500K daylight,
warm inviting mood. The mug is in sharp focus with a slight depth of
field falloff on the background. The mug is color-accurate to the
original. Photorealistic, 1080x1080, no watermark, no text.
A lifestyle photograph of a leather crossbody bag on a woman walking
through a sunlit urban street, the bag is the focal point in sharp
focus, natural daylight from above-right, color temperature 5500K,
energetic mood. The bag's leather texture, color, and hardware are
preserved exactly. Photorealistic, 1080x1350 portrait, no watermark,
no text, no fake social handles.
Why it works: the context is named, the lighting is specific, the focal point is the product, the color-accuracy rule is explicit.
The common failure mode: the model de-prioritizes the product in favor of the scene. The fix: "The product is the focal point, in sharp focus, color-accurate to the original." Repeat the color-accuracy rule. The product must dominate the frame.
Pattern 3: Detail / close-up shot (texture, material, craftsmanship)
The "show the quality" prompt. A close-up of the product's most distinctive feature — the stitching on a leather bag, the glaze on a ceramic mug, the texture of a fabric, the mechanism of a watch.
The prompt structure:
An extreme close-up product photograph of [PRODUCT], focused on
[DETAIL FEATURE]. The detail fills 70% of the frame. Macro lens
(100mm), f/4, shallow depth of field. Studio lighting that flatters
the [material]: [specific lighting for the material]. The detail
is color-accurate to the original. Photorealistic, 1024x1024, no
watermark, no text.
Examples:
An extreme close-up of a leather crossbody bag, focused on the
hand-stitched seam along the strap. The stitching fills 60% of the
frame. Macro lens (100mm), f/4, shallow depth of field. Soft
directional lighting from camera-left at 30 degrees, raking light
to emphasize the thread texture. The leather color (#8B4513 saddle
brown) and thread color (#F5DEB3 wheat) are accurate. Photorealistic,
1024x1024, no watermark.
An extreme close-up of a ceramic coffee mug, focused on the
hand-thrown glaze texture on the exterior. The glaze fills 60% of
the frame. Macro lens (100mm), f/5.6, shallow depth of field. Soft
diffused lighting from above to show the glaze's variation. The
glaze color (#2F2F2F charcoal) and texture (matte with subtle
crystalline variation) are accurate. Photorealistic, 1024x1024, no
watermark.
Why it works: the detail feature is named explicitly, the macro lens setup is specific, the lighting is tuned to the material, the colors are in hex.
Pattern 4: On-model / in-use shot (apparel, accessories, jewelry)
The "show the product on a person" prompt. The product is on a model, in a real-world context. The product is preserved; the model is supporting.
The prompt structure:
A photograph of a model wearing/using [PRODUCT] in [CONTEXT]. The
model is [description: age, ethnicity, body type, expression]. The
product is in sharp focus, color-accurate, and preserved exactly as
designed (color, pattern, fit, material). The model is secondary to
the product. [Lighting setup]. Photorealistic, [aspect ratio], no
watermark, no text, no fake brand marks.
On Nano Banana 2 specifically: the multi-image reference pattern works here. Pass the product photo as Reference 1, the model's pose/setting as Reference 2, and specify "preserve the product from reference 1, place on the model from reference 2."
On GPT Image 2 specifically: strong at virtual try-on. Use the multi-image editing pattern from the multi-image post.
On Seedream 5.0 Lite specifically: the reasoning model is good at preserving the product while changing the model/scene.
The common failure mode: the model becomes the focus and the product becomes secondary. The fix: "The product is in sharp focus, color-accurate, and preserved exactly. The model is secondary to the product."
Pattern 5: Scale / size reference shot (showing product size)
The "show how big it is" prompt. The product is shown with a recognizable size reference — a hand, a coffee cup, a phone, a ruler.
The prompt structure:
A product photograph of [PRODUCT] shown with [SIZE REFERENCE] for
scale. The product and the size reference are both in sharp focus.
The product is color-accurate and preserved exactly. Studio lighting,
no shadows, [aspect ratio], no watermark, no text.
Examples:
A product photograph of a small leather cardholder, shown next to
a standard credit card (8.56cm x 5.4cm) for scale. Both are in
sharp focus, on a clean white background, even studio lighting,
no shadows. The cardholder is color-accurate (#1A1A1A black).
1024x1024, no watermark, no text.
The common failure mode: the model gets the size wrong (makes a "small" cardholder look like a wallet). The fix: name the actual dimensions in the prompt and include the size reference.
Pattern 6: Group / variant shot (showing multiple products in one image)
The "show the product line" prompt. Multiple products in one image — a set, a collection, a family of variants. Used for "shop the look," gift guides, product family pages.
The prompt structure:
A product photograph of [PRODUCT SET] arranged on [SURFACE]. Each
product is in sharp focus, color-accurate, and preserved exactly.
Even studio lighting, no hard shadows. [Layout description]. The
products are spaced [X] cm apart, viewed from [angle]. Photorealistic,
[aspect ratio], no watermark, no text, no fake brand marks.
On Nano Banana 2 specifically: the only model that handles 5+ product references in one call. Pass each product as a separate reference.
The common failure mode: the products blend together, or the model invents fake products. The fix: list each product by name, specify the layout (grid, row, cluster), and require "no invented products, no fake brand marks."
The lighting rules for product photography
The model needs to be told the lighting setup. Without it, the model picks a default that often does not flatter the product.
| Product type | Lighting setup | Why |
|---|---|---|
| Matte product (ceramic, fabric, paper) | Soft diffused light from camera-left, even fill from camera-right | Even illumination, no harsh highlights |
| Glossy product (glass, polished metal, lacquer) | Softbox key + black flags to control reflections | Controlled highlights, no blown-out reflections |
| Transparent product (glass, liquid) | Backlight + side fill, dark background | Shows the product through transparency |
| Metallic product (jewelry, hardware) | Soft key + specular highlights, dark or gradient background | Shows the metal's luster |
| Textured product (leather, wood, stone) | Raking light from low angle | Emphasizes the texture |
| Food / organic product | Soft natural light from above, warm color temp | Natural, appetizing look |
| Small product (jewelry, electronics) | Macro ring light or twin flash | Even illumination at close range |
The general rule: match the lighting to the material. Matte wants soft, even light. Glossy wants controlled highlights. Transparent wants backlight. Textured wants raking light.
The color-accuracy rules
Five rules that make the colors right:
- Specify the color in hex if possible. "#8B4513 saddle brown" not "brown."
- Specify the finish. "matte," "glossy," "satin," "metallic," "transparent."
- Specify the material. "ceramic," "leather," "glass," "fabric," "metal." The model maps material to default rendering.
- Reference a real product when possible. "Like an Apple Magic Mouse in space gray" gives the model a known target.
- Generate against a known background. Pure white or pure gray gives the model a color reference. Colored backgrounds shift the product's apparent color.
The batch-generation workflow
For 50+ SKUs in a product line, the workflow is:
Step 1: Define the style guide once.
- Background: pure white #FFFFFF
- Lighting: softbox key from camera-left 45 degrees, fill from camera-right
- Camera: straight-on, product fills 80% frame
- Output: 1024x1024 PNG, no watermark
- Color: color-accurate to source
Step 2: Build a prompt template.
A studio product photograph of [PRODUCT_NAME] in [COLOR_HEX]
[FINISH] [MATERIAL], centered on a pure white background
(RGB 255, 255, 255). Soft even studio lighting with a large
softbox key from camera-left at 45 degrees and a smaller fill
from camera-right. No hard shadows. The product fills 80% of
the frame. Color-accurate to the original product. Photorealistic,
1024x1024, no watermark, no text, no logos, no people.
Step 3: Generate one product, verify, lock the style.
Generate the first product. Check: white background pure? Lighting consistent? Color accurate? Composition centered? Once it matches, lock the seed (if the model supports it) and the prompt template.
Step 4: Batch-generate the rest.
For each subsequent product, change only the product description (name, color, finish, material). Keep the lighting, background, camera, and aspect ratio identical. The result is 50 products that all look like they came from the same studio.
Step 5: QA the batch.
- Open all 50 in a grid view
- Check for consistency: same background, same lighting direction, same composition
- Flag any product that drifts on color, material, or background
- Regenerate the flagged ones with a more explicit prompt
The batch workflow is what makes 50 SKUs shippable in a day instead of a week.
The per-platform checklist
| Platform | Background | Aspect ratio | Min size | Special rules |
|---|---|---|---|---|
| Amazon | Pure white #FFFFFF | 1:1 | 1000x1000 (2000x2000 recommended) | No text, no logos, no badges, no watermarks, no props |
| Shopify | White or lifestyle | 1:1 | 2048x2048 | Soft shadow OK, lifestyle OK |
| Etsy | White or lifestyle | 1:1 or 4:5 | 2000x2000 | Lifestyle OK, props OK |
| Instagram Shop | Lifestyle preferred | 1:1 or 4:5 | 1080x1080 | Lifestyle, lifestyle, lifestyle |
| Lifestyle | 2:3 | 1000x1500 | Vertical, lifestyle, scrollable | |
| Facebook Shop | White or lifestyle | 1:1 | 1024x1024 | Lifestyle OK |
| Google Shopping | Pure white #FFFFFF | 1:1 | 100x100 (800x800 recommended) | No text, no logos |
| Email (Klaviyo, etc.) | Lifestyle | 1:1 or 4:5 | 600x600 | Lifestyle, lifestyle, lifestyle |
The platform rule: Amazon and Google Shopping are the strictest (pure white, no text, no logos). Instagram and Pinterest want lifestyle. Shopify and Etsy allow both. Pick the platform first, then the prompt.
The pre-flight checklist
Before you ship a product image:
- Background matches platform requirements (pure white for Amazon, lifestyle for Instagram)
- Color is accurate to the actual product (verify against the source)
- Material is rendered correctly (matte looks matte, glossy looks glossy)
- Lighting flatters the material (soft for matte, controlled for glossy)
- Composition is consistent across the batch (same angle, same framing, same product size in frame)
- No fake text (no fake brand marks, no fake labels, no fake URLs)
- No fake logos (no invented brand marks, no fake TM/®)
- No watermark (unless intentional)
- No people (unless it's an on-model shot)
- Aspect ratio matches the platform (1:1 for Amazon, 2:3 for Pinterest, etc.)
- Size meets the platform minimum (1000x1000 for Amazon, 2048x2048 for Shopify)
- Verified at the target size (open the actual listing and look)
Skip any of these and the product image looks "AI-generated" instead of "professionally photographed."
The summary
Product photography is the most commercial use of AI image generation. The 6 patterns — pure white studio, lifestyle, detail, on-model, scale reference, group — cover ~90% of ecommerce needs.
- Pick the model for the asset type. NB2 for batch studio shots, GPT Image 2 for text-heavy or multi-image composites, Seedream 5.0 Lite for reasoning over material accuracy, Midjourney v7 for lifestyle/stylized.
- Use the lighting rules — match the lighting to the material (matte = soft, glossy = controlled, textured = raking).
- Use the color-accuracy rules — hex codes, named finishes, named materials, real-product references.
- Use the batch workflow — define the style once, build a prompt template, generate 50 SKUs that all look like the same studio.
- Use the per-platform checklist — Amazon and Google Shopping want pure white, Instagram and Pinterest want lifestyle, Shopify and Etsy allow both.
The model is not the problem. The prompt is underspecified, the platform rules are not named, the lighting is not tuned to the material, and the batch is not consistent. Add the structure, the rules, the workflow, and product photography becomes a reliable, scalable production pipeline.
