Product Photography Prompt Expert
Studio-grade product photo prompts that keep your real product accurate
For e-commerce sellers, brand and marketing teams, and freelancers who create product visuals with AI image models.
Generic prompts produce product images that look generated - and generators quietly change labels, logos, colors and proportions. This skill makes your agent work like a product photographer: it picks the shot type for the channel (white packshot, hero, lifestyle, in-hand scale, flat lay, macro, levitation, splash, group, seasonal campaign), chooses the lighting setup by material (dark-field glass, gradient reflections for chrome, raking light for fabric, backlight for drinks), and specifies surface, props, lens and color grade.
When you have a photo of your real product, it chooses between editing the photo, reference-guided generation and compositing, writes preservation instructions, lists what typically drifts, and gives you a verification checklist - so you never publish an image that misrepresents what you ship.
It includes main-image rules for Amazon (pure white, ~85% fill, 1000 px minimum, 1600+ recommended) and Walmart (white, 1:1, 2200 px recommended), plus Amazon's 2026 requirement to tag photorealistic AI-generated people with metadata - verified from official seller sources in October 2026, with a reminder to re-check your seller account.
You get: decision tables, a 10-shot library, lighting and material playbook, set/prop/grading and seasonal kits, model adapters for OpenAI GPT Image 2.5, Google Nano Banana, ByteDance Seedream 5.0, FLUX 3 / FLUX.2 and Midjourney V8, five worked examples, a self-check list and a size-fitting script with marketplace presets. Marketplace acceptance is never guaranteed; always review outputs before publishing.
Highlights
- Packshots, hero, lifestyle, flat lay, macro, levitation, splash, seasonal
- Lighting chosen by material: glass, chrome, fabric, food, cosmetics
- Reference-photo protocol that flags label and logo drift
- Amazon and Walmart main-image rules plus AI-people tagging
- Adapters for GPT Image 2.5, Nano Banana, Seedream 5, FLUX 3, Midjourney
Example
Here's a photo of my sage-green water bottle on my kitchen counter. Make it my Amazon main image.
**Shot plan:** White packshot | Amazon main | 1:1, 1920x1920 | Path A (edit the real photo - never generate a new bottle) **Model:** gpt-image-2.5-sunburst (edit) **Settings:** size "1920x1920", quality "high", output_format "png" **Prompt** Use: Amazon main product image, square. Image 1 is the exact product. Keep the bottle's shape, proportions, matte sage-green finish, brushed-steel cap and the printed logo and text exactly as in image 1. Change only the background and lighting: seamless pure white background (RGB 255, 255, 255), bottle centered, filling about 85 percent of the frame height, slight three-quarter turn, even high-key light from two large softboxes, a very faint contact shadow under the base. No added text, logos, props or watermarks. **Fidelity check:** compare logo and text letter by letter; color-pick the green; 50% overlay for proportions. **Compliance:** every background pixel 255/255/255; whole product visible; >=1600 px long edge.
Instagram ad for my faceted glass perfume with a gold cap. Very luxury, dark mood, space for a headline.
**Shot plan:** Hero | Instagram feed | 4:5 | Path B (reference-guided, verify, composite label if it drifts)
**Model:** FLUX 3 Image **Settings:** aspect_ratio "4:5", resolution "2k", images [bottle.jpg]
**Prompt**
A luxury fragrance advertising photograph. The exact perfume bottle from image 1 - faceted clear glass, gold cap, label unchanged - stands as <bottle> on <slab>. <space> stays dark and empty for a headline. Dark-field lighting: two narrow strip lights behind the bottle drawing bright outlines along every facet, a warm glow through the amber liquid, a soft reflection on polished black marble. Palette black #0B0B0C, gold #C9A35B, amber #B5651D.
[{"id": "space", "bbox": [0, 0, 380, 1000], ...}, {"id": "bottle", ...}, {"id": "slab", ...}]
**Fidelity check:** facet count, cap shape, label text, fill level.Two gallery images for my 30 ml serum bottle: one on a bathroom shelf, one in a hand to show the size.
**Shot plan:** Lifestyle 4:5 + in-hand scale 1:1 | Amazon gallery | Path B **Model:** gemini-nano-banana-2.1 **Settings:** aspect_ratio "4:5" / "1:1", image_size "2K" **Prompt 1** Create a vertical 4:5 product gallery photograph... Using the provided photo as the exact product, keep the white 30 ml dropper bottle, its bulb and the blue label text exactly as they are. Place it on a pale travertine shelf beside a folded cotton towel, eucalyptus softly out of focus, soft morning window light from the left... **Compliance:** image 2 shows a photorealistic AI-generated hand, so Amazon requires the keyword contains-synthetic-performer in the file's XMP dc:subject before upload (or shoot a real hand). Styling props are not included items - keep them secondary.
What's inside
product-photography-prompt-expert/ ├── agents/ │ └── openai.yaml ├── references/ │ ├── lighting-and-materials.md │ ├── marketplace-specs.md │ ├── model-adapters.md │ ├── reference-photo-fidelity.md │ ├── sets-props-grading.md │ ├── shot-types.md │ └── worked-examples.md ├── scripts/ │ └── aspect_fit.py ├── LICENSE.txt ├── README.md └── SKILL.md
Install by unzipping into your agent's skills folder. Install guide →