Quick Answer

To photograph transparent products like glass bottles, perfumes, and liquids, you control the optics around the bottle, not the bottle itself. The studio path uses bright-field or dark-field lighting, polarizers, dulling spray, and deliberate edge definition. The production path, used by modern DTC brands managing dozens of SKUs, anchors a clean reference photo of the real bottle inside an AI workflow such as Nightjar so refraction, label, and fill level stay locked while lighting, background, and aspect ratio change. Generic prompt-only AI tools fail on glass because they have no internal scene graph to reason about refraction. Reference-anchored tools do.

Why Transparent Products Are the Hardest Category in Product Photography

A bottle is not a subject. It is a lens. The camera does not photograph the glass. It photographs whatever the glass is showing it: the lights, the studio walls, the photographer, the ceiling. That is why a clean perfume bottle on a white sweep often photographs as a faint outline with no shape. Glass on white is white through white. There is nothing for the camera to find unless edges are deliberately added by lighting. Anyone searching for how to photograph transparent products is really searching for how to control everything around them.

The cost of getting this wrong is unusually steep. 77% of shoppers say image quality is important to their purchase decisions, and 67% rate image quality higher than product descriptions or customer reviews. Transparent products carry a second penalty most categories do not: returns. 71% of consumers have returned products because the actual item did not match what was shown online. Misread a perfume bottle's color, fill level, or label, and the buyer feels misled the moment they open the box.

There is also a brand-side problem the existing literature mostly skips. A working photographer's question is "how do I shoot this one bottle?" A brand operator's question is "how do I produce 14 perfumes, 6 oils, or 20 beverage SKUs that read as one shoot, then refresh them next season?" Those are different questions, and they need different tools. This guide is split into two tracks. Track one is the studio path, written for the purist. Track two is the catalog production path, written for the operator.

The Physics You Are Actually Fighting

Three optical effects make glass photography hard. They are worth naming, because every studio technique and every AI failure traces back to one of them.

Refraction. Light bends as it passes through glass and liquid. A bottle should display a refracted version of whatever sits behind it. Move the bottle one inch and the refraction shifts. Change the liquid's color or density and the refractive index shifts. The fill line is its own optical event, with the upper air-volume refracting differently from the liquid below.

Reflection. Every shiny surface in the studio prints itself onto the bottle. The softbox above prints on the shoulder. The white wall behind the photographer prints on the cap. The floor prints on the base. A bottle that looks clean to the eye looks like a mirror to the lens.

Edge loss. A glass bottle on a flat white sweep contains no edges to find. Edges have to be added by lighting, not discovered by the camera. This is the single most counterintuitive part of the craft and the reason most amateur glass shots look ghostly.

For tinted liquids, all three effects compound. Amber whisky refracts differently from clear gin. Green olive oil reads as a different optical body than a clear neroli essence. Photographing six liquid color variants is not really photographing one product six times. It is photographing six different lenses.

Track One: The Studio Path

A serious glass setup runs on two lighting traditions, plus a small library of physical techniques that exist for one reason: to control reflections.

Bright-Field Lighting (Light Background, Dark Edges)

The background is lit. The body of the glass appears bright. Edges are defined by black flags or cards that act as negative fill. "The background is lit and the body of the glass appears white, with edges defined by dark cards that create negative light".

Best for: clean catalog white-background listings, Amazon main images, anywhere the bottle has to read as light through light.

Setup:

• Backlit frosted acrylic sheet roughly 1.2m by 1.2m, lit evenly from behind by one or two strobes through diffusion.
• Key softbox front-side, slightly above center.
• Two black flags or cards on either side of the bottle, just out of frame, to absorb light at the edges and print dark outlines on the glass.
• Optional fill card on the opposite side of the key, kept dim.

Dark-Field Lighting (Dark Background, Bright Edges)

The background reads black. The bottle's edges are lit and glow. "The black background is seen through the body of the bottle, but the edges are defined by the light that goes behind the black card".

Best for: dramatic perfume, premium spirits, hero shots that have to feel expensive without being shouty.

Setup:

• Black background, matte and clean.
• Two strip softboxes raking the bottle from behind on either side.
• A black card directly behind the bottle to keep the body dark, so only the edges catch the strip light.
• Camera slightly below shoulder height to elongate the bottle.

Polarization, Dulling Spray, and Killing Hot Spots

Polarizing filters on lights and lens cut studio reflections. Dulling spray (a matte aerosol) handles stubborn hot spots on shoulders, caps, and atomizers. Black flags above the bottle absorb stray light from the ceiling. Together these tools exist to turn the studio invisible to the bottle.

Bottle Prep, Cleaning, and Handling

Wipe the bottle with 99% isopropyl alcohol, not 70%, which leaves smears and dries slow. Work in cotton or nitrile gloves. Inspect the bottle under raking light from multiple angles before every frame. A single fingerprint at the base will print clearly under bright-field light.

Faking Condensation (Beverage and Chilled Bottles)

Mix 1:1 glycerin and water in a fine-mist sprayer; the spritz holds for hours, sometimes days, where plain water slides off in minutes. For larger heroic droplets, dab glycerin with a small brush and let surface tension do the work.

Label Visibility on Curved Bottles

Side fill plus rim lighting flattens the curve enough for the label to read. The camera angle slightly above center reduces label distortion. For embossed or transparent labels, an oblique key light catches the relief and gives the typography somewhere to land.

A working glass kit therefore looks like this: key light, fill light, edge or rim light, background light, polarizers for lights and lens, dulling spray, frosted acrylic diffusion panel, black flags and cards, gloves, 99% isopropyl alcohol, and a fine-mist sprayer with glycerin solution.

The Honest Cost of Doing This Well

The studio path produces beautiful results. It is also expensive in a way that most guides quietly skip past. Glass and beverage photography typically runs $500 to $2,000 per setup, and a multi-angle plus lifestyle treatment for a single SKU runs $3,000 to $5,000. Day rates in 2025 range from $800 to $5,000 and up, averaging $2,000 to $3,000. Retouching adds $50 to $300 per image.

Worked Example: A 14-SKU Perfume Catalog Refresh

A boutique fragrance brand launching a seasonal capsule with 14 perfumes wants four image variants per SKU: a white-background listing, a marble counter lifestyle, a bathroom shelf lifestyle, and a 9:16 social vertical.

• 14 SKUs x 4 image types x a conservative $750 per setup = $42,000 at the low end.
• Retouching: 56 final images x $50 to $150 = $2,800 to $8,400.
• Realistic full catalog production, including studio rental and props: $45,000 to $90,000 for one seasonal refresh.
• Timeline: 4 to 8 weeks including reshoots.

The math derives from ProShot Media and Mark Mendoza pricing applied to a representative DTC perfume catalog of 14 SKUs.

The deeper point sits underneath the number. The consistency tax on transparent products is the largest hidden line item in DTC fragrance and beverage marketing. Most guides describe how to shoot one bottle. The brand-side problem is producing the next thirteen that have to match, and the next capsule after that.

Track Two: Why Generic AI Tools Fail at Glass

Generic AI tools fail at glass because diffusion models trained on broad image corpora treat reflections and refractions as flat 2D texture, not as the optical consequence of a 3D scene. The model has no internal scene graph to know that the bottle should be showing a refracted version of the marble counter behind it. That is the technical reason ChatGPT, Midjourney, and raw SDXL outputs fall apart on transparency, even when the rest of the scene looks photographic.

The Four Failure Modes of Prompt-Only AI on Transparent Products

1. Hallucinated liquid contents. The user uploads a reference of an amber whisky. The output shows a pale rosé. The model invents a color and fill level that has nothing to do with the source bottle.
2. Drifting labels. Embossed and printed labels warp, change typography, or vanish entirely between Generations. Two outputs of the same prompt produce two different brand names.
3. Broken refraction. The bottle does not refract the actual background scene. A bottle placed on a real beach reference still refracts a generic sky, because the model has no concept of "what the bottle should be seeing."
4. Reflection confusion. The model cannot distinguish the studio softbox reflection that should be removed from the internal refraction that gives the bottle its 3D shape.

Why This Happens (Technical, Not Marketing)

"AI interprets visual data, not physical properties; it sees the reflections and refractions of a wine glass as solid geometry". "Diffusion models do not maintain internal 3D scene graphs. They have no concept of camera position, vanishing points, or occlusion relationships between subject and reflection". And critically: "An AI cannot inherently tell the difference between an unwanted studio softbox reflection and the necessary internal refraction that gives a glass object its three-dimensional shape".

A controlled, ingredient-based AI workflow with the real product image anchored as a reference is the only AI approach that respects glass physics in a repeatable way. The model is no longer asked to invent a bottle. It is asked to relight a bottle that already exists.

How AI Tools Compare for Transparent Product Photography

Cost figures reflect 2025 to 2026 published rates.

How a Controlled AI Workflow Handles Glass

The workflow below is the production model most DTC brands now use for their second, third, and fourth catalog refreshes, after they have one clean studio reference per SKU in hand.

Step 1: Anchor the Real Bottle

Upload one clean reference photo per SKU. Product preservation locks the real label, fill level, color, and silhouette. Generic models invent. Reference-anchored tools preserve. This single decision (the input is the real bottle, not a description of one) is what separates production-grade AI from prompt-only tools.

Step 2: Define the Refraction Language Once

Build, or pick from a library of curated styles, a Photography Style for bright-field listing language, dark-field hero language, or lifestyle bathroom counter language. The Style encodes lighting direction, edge definition, reflection softness, and color temperature. The same Style applied across 14 bottles produces 14 images that read as one shoot. See how to maintain a consistent aesthetic across all your AI images.

Step 3: Lock the Frame with a Composition

A Composition sets bottle scale, framing, camera elevation, and crop. This is what catalog grids need: every bottle occupies the same percentage of the frame, sits at the same height, and faces the same direction. It also meets Amazon's 85% frame coverage requirement reliably across the full SKU line, which is hard to guarantee freehand.

Step 4: Place the Bottle in a Real Scene, Not a Generic One

Use the Edit tab with @image1 for the bottle and @image2 for the scene reference (a real marble counter shot, a real bathroom shelf, a real beach plate). The model receives both inputs explicitly, so refraction has the correct reference instead of inventing a different scene. See how to blend a product into a stock photo or background image realistically.

Step 5: Save the Recipe, Reuse Across the Line

A Recipe locks the Photography Style, Composition, Background, aspect ratio, resolution, and output format. Apply the Recipe to the next bottle in one action. The next seasonal capsule reuses the same Recipes. Catalog production becomes a maintenance task, not a reshoot. See how to make AI product photos more consistent.

Step 6: Handle Edge Cases

• Liquid color variant. Recolor Edit Shortcut plus /color #2E5C3D shifts an amber oil to deep green while preserving label and fill line. See how to change the color of a product using AI.
• 9:16 ad cut from a 1:1 hero. Reframe Edit Shortcut plus /ratio 9:16. No reshoot.
• Marketplace zoom. Upscale to 2K or 4K, preserving label legibility and fill level rather than reinterpreting them. See how to make AI product photos look more professional.

Reusable Photography Styles solve the consistency problem that breaks AI glass photography. Without them, every Generation interprets refraction differently, and the catalog drifts within a week.

Marketplace Compliance for Transparent Products

Compliance is where most DTC photography efforts actually fail, regardless of whether the images came from a studio or a model. Glass amplifies the problem because edge loss and refraction make the rules harder to hit by eye.

• Amazon main image: pure white background at exact RGB 255, 255, 255; product at 85% or more frame coverage; long edge minimum 1,000 px (1,600 px recommended for zoom).
• Shopify and Etsy listings: 1:1 aspect ratio is preferred for grids. 4:5 for Instagram feed. 9:16 for stories and TikTok. 16:9 for hero banners.
• Zoom expectations: 2K and 4K outputs preserve label and fill level. Lower-resolution upscalers tend to reinterpret labels, which on glass usually means warped typography and shifted refraction.

A reference-anchored AI workflow exposes all of these as explicit settings: Background hex, Composition coverage, Aspect Ratio, Resolution, and output format. They become parameters of a Recipe, not judgment calls per image.

When to Shoot Studio, When to Use AI, When to Combine

There is no honest version of this article that says studio is dead. The honest version says studio and AI now solve different parts of the same problem.

Shoot studio when launching a brand new hero campaign that needs physical art direction, or when the product itself is in motion (pour shots, splashes, smoke, ice). A real photographer reading real refraction in real time is still the right tool for these.

Use AI when producing catalog-scale variants, lifestyle scenes across multiple backgrounds, color variants of one bottle, multi-format ad cuts, or seasonal refreshes that have to match a prior shoot. The cost curve here is not even close.

Combine when the goal is both. Shoot one clean studio reference per SKU. Then expand into the full catalog with reference-anchored AI. This is the production model most modern DTC fragrance, oil, and beverage brands now use, because it gets the physical art direction of studio and the catalog economics of software in the same workflow.

Frequently Asked Questions

How do you photograph clear glass bottles without reflections? Control the studio environment, not the bottle. Use a darkened shooting bay so ceilings, lights, and stands do not appear in the glass. Add polarizing filters on both the lights and the lens. Apply dulling spray to stubborn hot spots on shoulders and caps. For repeat catalog work, a reference-anchored AI tool such as Nightjar removes the studio-environment problem entirely, because reflections become deliberate lighting language inside a reusable Photography Style instead of stray gear in the room.

What lighting is best for transparent product photography? Bright-field for clean white-background listings (lit background, dark cards for edge definition) and dark-field for premium hero shots (dark background, rim-lit edges). The choice depends on whether the bottle should read as light through light with crisp outlines, or dark with glowing edges.

How do you make liquids look good in product photos? Backlight the liquid so the color reads through the bottle. For chilled or beverage shots, mist with a 1:1 glycerin-and-water spray for condensation that holds for hours instead of minutes. For tinted variants, either dye and refill, or use an AI Recolor workflow that preserves the bottle silhouette and label while swapping liquid color.

How do you photograph glass on a white background? Use bright-field lighting. Light the white background separately from the bottle. Place black flags or cards on either side of the bottle as negative fill, which prints dark edge lines on the otherwise-invisible glass. For Amazon, the background needs to be exact RGB 255, 255, 255 and the bottle has to cover at least 85% of the frame.

Why do my glass product photos look flat or invisible? Because the glass has no edges to find against a flat exposure. Glass on white is white through white. The fix is to add edge definition deliberately: black negative-fill cards in bright-field, or rim lighting in dark-field. Without one or the other, the camera has nothing to record.

How do you create condensation on bottles for photos? Mix glycerin and water in a 1:1 ratio and apply with a fine-mist sprayer. Glycerin holds the water on the surface for hours. For larger heroic droplets, dab glycerin with a small brush and let surface tension shape them.

Can AI generate realistic photos of glass and transparent products? Prompt-only tools (ChatGPT, Midjourney, raw SDXL) generally cannot, because diffusion models have no internal 3D scene graph and treat refraction as 2D texture. Reference-anchored tools that take the real bottle as input, such as Nightjar, preserve the actual label, fill level, and silhouette while regenerating only the lighting and background. That is the workflow that produces realistic, catalog-consistent glass at scale in 2026.

References

• Nightjar - Reference-anchored AI product photography platform
• Mark Mendoza Photography - Glass and beverage photography pricing
• ProShot Media - 2026 day rates
• Seller Labs - Amazon image requirements
• Westcott University - Dark-field lighting reference
• Use Your Noodles - Glycerin condensation technique
• Stelava - Bottle prep guide
• Imagework India - Generative fill artifacts on glass
• Development Corporate - Why AI struggles with the wine glass
• Grabon - Product photography statistics