AI Product Photography Prompts: What 1,081 Real Briefs Have in Common
The floating-bottle formula, the lighting and camera terms that appear most, and how three models get briefed differently — counted across 1,081 product prompts.
On this page
What 1,081 product prompts have in common
Most guides to AI product photography are one person's favourite prompt. This one is a count. We took every published item in the LocalBanana gallery tagged as product, packaging, advertising or e-commerce work — 1,081 prompts out of 11,763 with a usable prompt, as of 7 September 2026 — and measured what they actually ask for: which products, which lighting, which camera vocabulary, and how the three main models get briefed differently.
1,081
9.2% of the 11,763-prompt gallery corpus
123
site-wide median is 93 — product briefs run a third longer
17.5%
the single most common staging move after 'reflection'
Which products people actually shoot
The corpus is not a random sample of the world's products. It is what people bother to prompt for, and two categories dominate.
| Product category | Share of product prompts |
|---|---|
| Beverage (bottle, can, coffee, juice) | 22.0% |
| Skincare and cosmetics | 20.8% |
| Food and snacks | 10.5% |
| Sneakers and shoes | 8.6% |
| Bags and accessories | 8.2% |
| Electronics (earbuds, phones, watches) | 7.7% |
| Furniture | 6.0% |
| Perfume | 3.9% |
Beverages and skincare together are more than two fifths of the corpus, and they share a physical property: they are mostly glass, liquid and condensation. That explains a lot of the staging vocabulary in the next section — the techniques people converge on are the ones that flatter a wet bottle.
What the prompts are for is just as lopsided. 54.5% describe an advertisement, poster or campaign image. Only 3.2% mention an e-commerce listing, main image or detail page, and 5.0% a social post. People are prompting for the hero shot, not the white-background catalogue frame.
The floating-bottle formula
Strip the adjectives out of a thousand product prompts and a small set of staging moves remains. These are the ones that appear most, with the share of prompts that use each:
| Staging move | Share of product prompts |
|---|---|
| Reflection (mirror surface, reflective floor) | 18.2% |
| Levitating or floating product | 17.5% |
| Smoke, mist, fog or haze | 10.5% |
| Macro detail | 9.5% |
| Splash or water surface | 9.3% |
| Condensation droplets | 6.8% |
| Exploded view | 0.8% |
| Caustics | 0.6% |
Put the top five together and you have the image that a large share of the corpus is trying to make: a bottle suspended above a reflective surface, wrapped in mist, with a frozen splash and droplets on the glass. It works because it solves three problems at once — a floating object has no awkward contact shadow, a splash fills the empty space around a small product, and condensation gives the model a texture to render instead of a flat plastic surface.
The formula in one prompt: suspended bottle, condensation on every bead, frozen splash, whole fruit and ice for scale, a gradient background matched to the liquid, and a slight tilt to read as motion. 243 words, and nothing in it is decorative.
Create an ultra-realistic 3D commercial-style product shot of a premium cherry juice bottle, suspended mid-air with intricate condensation droplets on its surface. The bottle should appear fresh and vibrant, with each condensation bead reflecting ambient light to enhance photorealism. Surround the product with dynamic elements like splashing droplets of cherry juice, whole cherries, and ice cubes, frozen in high-speed motion, each element sharply defined with vibrant clarity. Floating cherry stems and leaves should also be included to enhance the sense of freshness and energy. Set the background against a rich, deep red and burgundy gradient, which complements the rich color of the cherry juice and evokes a sense of indulgence and premium quality. The product should be centrally placed, slightly tilted to convey a sense of movement and sophistication. Use cinematic, studio-style lighting with bright highlights reflecting off the bottle, crisp shadows, and high contrast to create a luxurious, polished look. Ensure the bottle’s label is clearly visible, with subtle reflections beneath it, adding depth and realism to the scene. The overall aesthetic must evoke indulgence, freshness, and premium quality, with all elements contributing to a high-end, visually striking image. The scene should feel rich, fresh, and full of vitality,with a focus on the vibrant color and fresh nature of the cherry juice. Technical Specifications: Aspect Ratio: 4:5 Resolution: Ultra-HD quality Lighting: Studio-style, cinematic with bright highlights, subtle reflections, and high contrast Detailing: Extreme attention to condensation, droplets, and high-speed motion of elements
Two things the table does not show. First, the rare moves are not rare because they fail: caustics (the light patterns water throws) appears in 0.6% of prompts and exploded view in 0.8%, and both produce distinctive images precisely because almost nobody asks. Second, the formula is a beverage formula. Skincare prompts use it too, but the highest-viewed skincare items in the corpus are the still ones — a dropper bottle on a wooden block with a long shadow, not a splash.
Lighting and background: what people ask for, and what they leave out
Lighting vocabulary in product prompts is narrower than in the corpus overall — and heavier on one term.
| Lighting term | Share of product prompts |
|---|---|
| studio lighting | 22.8% |
| rim light | 7.9% |
| natural or window light | 7.3% |
| neon | 5.9% |
| hard shadow or direct sunlight | 2.3% |
| golden hour | 2.1% |
| softbox | 1.9% |
| backlight | 1.4% |
Nearly a quarter of product prompts say studio lighting and stop there. It is the single most common lighting instruction, and it is close to meaningless — a studio can be lit with one bare bulb or twelve softboxes. Compare that with rim light at 7.9%, which actually tells the model where the light goes: behind the product, tracing the silhouette. Prompts that name a direction, a source or a quality of light are a minority, which is why they stand out.
Backgrounds are described even less often. white or seamless background appears in 8.1% of prompts, a gradient in 3.2%, marble in 2.5%, a pastel tone in 2.4%, a podium or plinth in 1.9%. Roughly four in five product prompts never say what the product is standing on or in front of — they describe the hero and let the model invent the room.
55 words, and every clause is a decision: mirror-like water surface, petals frozen mid-air, a pastel gradient, volumetric sunlight. Short prompts work when each phrase is a specification rather than an adjective.
High-end commercial shot of a minimalist glass perfume bottle filled with pale rose gold liquid. It is resting on a mirror-like water surface. Floating silk rose petals and morning dew droplets surround the bottle, frozen in mid-air. Soft pastel pink and white gradient background with dreamy volumetric sunlight. Elegant, ethereal, and romantic atmosphere, --ar 3:4
premium perfume bottle, luxury product photography, studio lighting, high quality, 8K
rectangular glass perfume bottle, brushed gold cap, centred on a black reflective acrylic surface, rim light from behind tracing the silhouette, soft key from camera left, dark charcoal gradient background, 100mm macro at f/8, condensation absent, label blank
Camera vocabulary: resolution words beat lens words
| Camera or composition term | Share of product prompts |
|---|---|
| 8K or 4K | 25.1% |
| depth of field or bokeh | 24.9% |
| centred composition | 16.0% |
| close-up | 11.1% |
| 85mm | 7.9% |
| top-down or flat lay | 7.4% |
| aperture (f/…) | 6.5% |
| low angle | 5.6% |
| eye level | 5.4% |
| three-quarter view | 3.5% |
| hero shot | 3.5% |
| 100mm | 1.7% |
| 45-degree angle | 1.2% |
One in four product prompts asks for 8K. It is the most common camera-adjacent term in the corpus and the least useful: the output resolution is set by the model route, not by the prompt. It survives because it feels like a quality instruction.
The shape of the table is the more useful finding. depth of field appears in a quarter of prompts, but only 6.5% name an aperture and only 1.7% name the 100mm macro lens that real product photographers reach for — 85mm, a portrait lens, outnumbers it more than four to one. Angles are mentioned in a minority of prompts, and when they are, low angle (5.6%) slightly beats eye level (5.4%). The classic 45-degree three-quarter product angle, the default of every catalogue shoot, appears in 1.2%.
If you want to be more specific than most of the corpus, you do not need to be clever. Name the angle, name the lens, name the aperture. Three short phrases put a prompt in the top tenth for camera specificity.
A rare angle in the corpus, stated plainly: low perspective looking up, product on the sharp edge of a glass podium, water running down both. Naming the camera position does more for the composition than any adjective in the prompt.
Dramatic low-angle product photography of the dark red Sokolov Beauty hand cream bottle standing on the sharp edge of a transparent glass podium. The bottle is shot from a low perspective looking up. Clear water is elegantly dripping and flowing down the sides of the bottle and across the glass surface. Several juicy, glossy dark red cherries are placed next to the bottle on the podium. Deep burgundy background with a soft light gradient fading toward the top. Dramatic cinematic lighting with strong highlights and deep shadows, creating a luxurious and sensual mood. Highly detailed water droplets and reflections, premium commercial beauty photography, sharp focus, 8K resolution, photorealistic.
Shooting your own product from a reference photo
10.8% of product prompts start from an uploaded photo — the words reference image, uploaded, attached or keep the packaging appear in one prompt in nine. Split by model it is 14.3% for Nano Banana Pro, 12.7% for GPT Image and 1.3% for Midjourney, which mirrors how well each route holds a real product's shape.
The reference-based prompts have a recognisable structure that the from-scratch prompts do not: they spend their first sentence on what must not change (shape, label, colours, proportions) and only then describe the scene. The most-viewed pattern in this group is the storyboard grid — one product, nine panels, each panel a different concept.
Nine concepts for one uploaded product in a single 3:4 frame: hero still life, macro texture, liquid interaction, floating elements, colour-driven scene. The visual rules section pins the product to 100% shape accuracy before any concept is described.
Create a 3×3 grid in 3:4 aspect ratio for a high-end commercial marketing campaign using the uploaded product as the central subject. Each frame must present a distinct visual concept while maintaining perfect product consistency across all nine images. Grid Concepts (one per cell): 1. Iconic hero still life with bold composition 2. Extreme macro detail highlighting material, surface, or texture 3. Dynamic liquid or particle interaction surrounding the product 4. Minimal sculptural arrangement with abstract forms 5. Floating elements composition suggesting lightness and innovation 6. Sensory close-up emphasizing tactility and realism 7. Color-driven conceptual scene inspired by the product palette 8. Ingredient or component abstraction (non-literal, symbolic) 9. Surreal yet elegant fusion scene combining realism and imagination Visual Rules: Product must remain 100% accurate in shape, proportions, label, typography, color, and branding No distortion, deformation, or redesign of the product Clean separation between product and background Lighting & Style: Soft, controlled studio lighting Subtle highlights, realistic shadows High dynamic range, ultra-sharp focus Editorial luxury advertising aesthetic Premium sensory marketing look Overall Feel: Modern, refined, visually cohesive High-end commercial campaign Designed for brand websites, social grids, and digital billboards Hyperreal, cinematic, polished, and aspirational
We tested how well the three models hold a real object through an edit in AI image editing prompts, tested; the short version is that the first sentence of the prompt — the constraint — matters more than the rest. For a listing rather than a campaign, the Etsy product image guide covers the white-background workflow that only 3.2% of this corpus bothers with.
The reference workflow taken further: one packaging photo turned into a design sheet — hero render, six technical views, fold lines, dimensions in millimetres and material callouts. GPT Image, 115 words.
Using the supplied reference image, produce a professional industrial-design packaging illustration sheet for the package (PACKAGE TYPE). Place a hero 3D render in the middle with photoreal materials, gentle studio lighting, and a polished commercial finish. Around it, lay out the technical views: front, side, top, bottom, an oblique angle, and a flat unfolded layout. Add line sketches showing the frame structure, fold/crease lines, seam details, and dimension arrows labeled in millimeters. Annotate materials and surface finishes (matte, gloss print, plastic, paper, glass, etc.) with handwritten callouts. Drop in color swatches, supporting product mockups, and soft cast shadows. Background: clean sketchbook paper. Style: photoreal render mixed with pencil-sketch overlays, modern design aesthetic, ultra-detailed, portfolio quality.
Label text: half the corpus mentions it, one model gets asked most
48.6% of product prompts mention a label, logo, text or typography. It is the sharpest model split in the data: 67.6% of GPT Image product prompts involve text, against 43.1% for Nano Banana Pro and 17.1% for Midjourney. People route the shots that need a legible label to the model they trust to render it, and describe the label in more detail when they do.
A second, smaller group goes the other way: 7.6% of prompts explicitly say no text, no logo or no watermark, rising to 11.2% for GPT Image. That is the unbranded-mockup workflow — generate the bottle, add the label in a design tool afterwards — and it produces some of the cleanest results in the corpus because the model is not spending effort on letterforms.
Two more habits worth stealing. 8.3% of prompts use placeholders like [BRAND NAME] or [PRODUCT]; they are templates written to be reused, and they are concentrated in Nano Banana Pro prompts (16.7%). And 11.7% specify a colour palette, sometimes as hex codes — three times as common among GPT Image prompts (17.3%) as among Nano Banana Pro prompts (8.4%).
Label text handled the way the successful prompts handle it: the exact string in quotes, the typography material named (gold on matte black), and nothing else asked of the type. 55 words.
Ultra-cinematic premium coffee bottle labeled "NOIR BREW", matte black glass with gold typography, floating upright amid swirling espresso waves and roasted coffee beans. Steam and mist curling around the bottle, dramatic low-key lighting with warm highlights and deep shadows, rich brown color grading, macro condensation details, cinematic depth of field, photorealistic, 8K, luxury branding aesthetic.
Which model renders a label most reliably is a separate, controlled question — see which AI image generator renders text best, where the same twelve typographic prompts went to each route.
Length and structure: product briefs run long, and JSON ones run longer
The median product prompt is 123 words, against 93 for the gallery as a whole. Split by model: GPT Image product prompts have a median of 184 words, Nano Banana Pro 119, Midjourney 57. The same product, briefed three ways, gets three times as many words on one route as on another.
11.3% of product prompts are written as structured JSON rather than prose — almost exactly the site-wide rate of 11.1%. Their median views are 41, against 13 for prose prompts. That is a smaller gap than the 11× we measured across the full corpus, and the sample of 122 JSON prompts is small enough that we would not build a rule on it. The mechanism is the same as always: a structured prompt cannot leave the surface, lighting or camera slot empty without the author noticing.
A structured product template at the long end of the corpus: bottle proportions, glass thickness, label typography, cap material and the behaviour of light are each given their own clause. The scent_profile key at the top is the only variable — everything else is fixed.
{
"scent_profile": "Floral",
"prompt": "Ultra high-end luxury perfume campaign image for a world-class prestige fragrance. A premium rectangular perfume bottle with perfectly balanced proportions typical of iconic global luxury brands, refined, substantial, and timeless, never tall, never slender, never vial-like. The bottle is crafted from thick, heavy, crystal-clear glass with softly rounded edges, visible weight, optical depth, and precise craftsmanship, filled with luminous liquid appropriate to a high-end perfume. A minimal ivory or warm off-white label is centered on the bottle, featuring ultra-clean modern sans-serif typography, extremely sparse text, and confident negative space. The cap is a champagne-gold or pale-gold brushed metal cylinder with precise machining and subtle, controlled reflections. The bottle floats calmly in space with quiet authority and confidence. Surrounding the bottle are exactly three distinct types of scent-related visual elements autonomously selected by the AI based on the provided scent_profile. These three element types represent different dimensions of the fragrance family, such as structural material forms, botanical or organic references, and atmospheric or sensory effects. All elements originate from the broader natural, botanical, and material context of the scent profile as understood in high-end perfumery. The elements are abstracted, symbolic, refined, and art-directed, never literal, never illustrative, never decorative. The elements are positioned close to and partially embracing the bottle, following its silhouette and contours rather than floating freely in space, creating a sense of attachment and intentional interaction with the product. The elements gently overlap the bottle edges in places, while maintaining clarity of the label and overall form. Spacing is controlled and elegant, with no bulky masses and no excessive gaps, forming a cohesive sculptural arrangement designed around the bottle. The background is a rich, noble, scent-matched tonal gradient with depth and luminosity, not dark or gloomy, but dense and refined. The background color harmonizes with the scent_profile. The background maintains medium-to-high brightness with elevated color density, subtle light falloff, and a refined vignette that enhances contrast without heaviness. Lighting is museum-quality studio lighting with a controlled warm key light, soft sculpting fill, and a precise rim light defining the bottle edges against the noble background. Glass refraction and reflections are perfect and premium, with no harsh highlights. Depth of field is moderate: the bottle and label are razor sharp, while decorative elements soften subtly with depth but remain clearly articulated. The overall mood is iconic, restrained, confident, luxurious, and unmistakably expensive, resembling a flagship global luxury perfume campaign.",
"negative_prompt": "pure white background, dull gray background, muddy brown, overly dark background, gloomy lighting, flat background, e-commerce lighting, floating decorations detached from the bottle, scattered elements, bulky clusters, cheap, low-end, mass-market, indie look, playful mood, cartoon style, plastic materials, thin glass, lightweight bottle, sample bottle, vial, test tube, fashion vial, visible branding text, logos, watermarks, noise, grain, oversharpening, blown highlights, distorted glass, warped label, fisheye perspective, people, hands, extra objects",
"aspect_ratio": "3:4",
"quality": "ultra high detail, photorealistic, luxury commercial retouching",
"style": "iconic luxury perfume photography, global prestige brand, cinematic, timeless, haute parfumerie"
}If you would rather not write JSON, the structured prompt templates article shows the labelled-slot version that reproduces most of the effect in plain text.
Which model people brief for product shots
| GPT Image | Nano Banana Pro | Midjourney | |
|---|---|---|---|
| Product prompts | 481 (44.5%) | 371 (34.3%) | 228 (21.1%) |
| Share of that model's own corpus | 11.8% | 10.8% | 5.4% |
| Median words | 184 | 119 | 57 |
| Written as JSON | 11.2% | 18.1% | 0.4% |
| Mention label, logo or text | 67.6% | 43.1% | 17.1% |
Say studio lighting | 23.9% | 29.6% | 9.6% |
| Ask for macro detail | 6.7% | 15.1% | 6.6% |
| Ask for condensation | 6.9% | 10.5% | 0.9% |
| Start from an uploaded product photo | 12.7% | 14.3% | 1.3% |
Use [PLACEHOLDER] template slots | 5.6% | 16.7% | 0.4% |
| Describe an ad, poster or campaign | 73.6% | 48.0% | 24.6% |
Product work is roughly one in nine of everything sent to GPT Image and to Nano Banana Pro, and one in twenty of everything sent to Midjourney. The two routes with reference-image support get the reusable, macro-heavy, condensation-heavy briefs; GPT Image gets the posters with copy on them; Midjourney gets short adjective lists. We ran the three routes head to head on fixed prompts in Same prompt, three AI image generators, and the split matches what people have evidently learned by trial: hold a real product and its label on the first two, mood-board on the third.
A brief you can copy
The prompts that combine the corpus's best habits share a shape. Fill the slots, delete the ones you do not need, and keep the order — product first, surface second, light third, camera last.
Product: [exact object: material, colour, cap or closure, label state — "label blank" or "label reads 'NOIR BREW' in gold serif"]
Position: [standing / lying / floating 5cm above the surface], [centred / offset left], [tilt or angle]
Surface and background: [black reflective acrylic / raw oak block / mirror-like water], [seamless gradient from X to Y / matte pastel wall]
Staging: [condensation on the upper third only / frozen splash rising from the base / two whole fruit for scale / none]
Light: key [soft, from camera left, 45° above], rim [from behind, tracing the silhouette], fill [none — let the shadow side fall to black]
Camera: [100mm macro / 85mm], [f/8 for full sharpness / f/2.8 for a soft base], [eye level / low angle looking up / 45° three-quarter]
Output: [3:4 vertical], [no text / exact label text only], [no humans, no hands]
Three notes on using it. Aspect ratio is worth stating: 67.7% of product prompts in the corpus are vertical (3:4 alone is 23.7%), and the aspect ratio guide covers what each ratio does to a centred product. Drop 8K; it costs words and buys nothing. And if the brief is for a real product, put the constraint sentence — what must not change — before any of this.
FAQ
What is the best prompt for AI product photography?
There is no single best prompt, but the ones that work share a structure: name the exact product and label state, the surface it sits on, the direction of the key and rim light, and the lens and angle. In a 1,081-prompt corpus, the vaguest common instruction was studio lighting (22.8% of prompts) and the most over-used was 8K (25.1%); replacing both with a light direction and a lens does more than any adjective.
How do I make an AI product photo look professional?
Control the three things a studio controls: the surface, the light and the angle. Say what the product stands on (only about one in five prompts does), give the light a direction and a rim, and name a camera position — eye level, low angle or 45-degree three-quarter. Floating the product above a reflective surface with condensation and a frozen splash is the corpus's most common formula, and it works, but it is also the most recognisable.
Can I use my own product photo with an AI image generator?
Yes. 10.8% of product prompts in the corpus start from an uploaded reference — 14.3% of Nano Banana Pro prompts and 12.7% of GPT Image prompts, but only 1.3% of Midjourney prompts. Write the constraint first: what must stay identical (shape, label, colours, proportions), then the scene. The 3×3 storyboard grid — one product, nine concepts — is the most-viewed pattern in this group.
Which AI image generator is best for product photography?
The corpus splits the work by job. GPT Image receives the most product prompts (44.5%) and 67.6% of them involve label text or ad copy. Nano Banana Pro prompts are the most reference-heavy (14.3% start from a photo) and the most macro- and condensation-heavy. Midjourney gets short mood briefs — median 57 words — and almost never a reference image. For a real product with a legible label, use one of the first two.
Should the product photo be square or vertical?
Vertical. 67.7% of product prompts in the corpus specify a vertical ratio — 3:4 is the most common at 23.7%, then 9:16 (17.9%), 4:5 (13.3%) and 2:3 (12.8%). Square is 18.6% and 16:9 only 8.6%. Match the ratio to where the image will run before you write the composition.
