Bottle product photography is the image set that sells a drink online: a white-background packshot, a label-legible front panel, a chilled or poured hero, and a lifestyle frame. Glass makes it technically harder than most categories, because a bottle reflects the room it was shot in. What makes it expensive is different. On a drink, the label is a regulated artifact and the picture is a claim.
Every guide ranking for this term opens on lighting. Kill the hotspot, use a stripbox, flag the sides in black, fake the condensation with glycerine. That advice is correct and it describes a photography problem.
It stops one step short of the thing that gets a drink image rejected. If you sell alcohol, an advertisement that shows a label must show your approved label. If you sell on a marketplace, the main image rules forbid most of what a styled drink shot contains. If you generate the frame with AI, the model’s default behaviour is to redraw the label, which breaks both. This guide covers the rules first, then the shot list, then where generative models actually fail on glass and liquid, then cost.
Key Takeaways
- The label is not a texture. For alcohol, a label shown in an advertisement must be a reproduction of an approved label. TTB lists label and COLA mismatch as a common violation it finds in the market.
- Generative models redraw labels by default. The correct workflow preserves the customer’s real label pixels and generates only the room around the bottle.
- A bottle reflects the studio. Most bottle retouching cost is the removal of the softbox, the ceiling and the photographer from the glass.
- Marketplace main images kill the styling. Amazon wants a pure white background at RGB 255, 255, 255 with the product filling roughly 85% of the frame and no props, so the ice, the garnish and the wet slate all belong in the secondary slots.
- Flavour extensions are the real budget line. A six-flavour range is six labels on one bottle shape, which is a reshoot in a studio and a re-run in a workflow.
- A drink SKU needs about seven frames. One packshot, one label panel, one scale reference, one chilled hero, one serve, one group shot, one social crop.
What is bottle product photography?
Bottle product photography is the production of commercial images of a bottled or canned product for a product page, a retail listing and paid social. It covers the white-background packshot that marketplaces require, the front-panel shot that makes the label readable at thumbnail size, and the styled frames that carry the brand. Glass is the hard surface because it is transparent, curved and reflective at the same time, so the lighting shapes the bottle by what it reflects rather than by what it illuminates.
The technique that separates a good bottle shot from a flat one is that you light the reflection, not the object. A bright surface placed to the side of the bottle appears in the glass as a clean vertical band, which reads as an edge and gives the shape volume. Black flags do the same job in reverse, drawing a dark edge that separates the bottle from a light background. Nearly everything else in bottle work is a variation on those two moves.
The frame above is a generated packshot, and it shows the single most useful habit in this category. The label is blank. A model that is not asked to invent lettering cannot get the lettering wrong, which is the whole reason the production order below puts the real label in last.
The label is a regulated surface, not a texture
This is the part the lighting guides skip, and it is the part that decides whether a drink image can ship.
The Alcohol and Tobacco Tax and Trade Bureau defines an advertisement broadly. It is “any written or verbal statement, illustration, or depiction which is in, or calculated to induce sales in, interstate or foreign commerce, or is disseminated by mail” (27 CFR 4.61, 5.232 and 7.232, ttb.gov, accessed August 2026). A product photo on a paid social placement is an advertisement under that definition. So is the image on a retailer’s page.
Two consequences follow, and both are listed under prohibited practices at 27 CFR 4.64, 5.235 and 7.235.
Statements inconsistent with labeling are prohibited, and TTB states the rule for pictures directly: a label depicted on an advertisement must be a reproduction of an approved label (ttb.gov, accessed August 2026). Approved means it carries a Certificate of Label Approval. A generated label, a cleaned-up label, a label with the alcohol content nudged for legibility, or a label composited from a different SKU is not a reproduction of the approved one.
TTB then names the failure explicitly in its own list of common issues found in the market: “labels depicted in ads don’t match the COLAs” (ttb.gov, accessed August 2026). That is the single sentence that should govern how a drinks brand uses generative tools. It is not a hypothetical risk, it is the agency describing what it already sees.
Advertisements also have to carry mandatory information, including the responsible advertiser’s name and address or contact details, and the class and type designation exactly as it appears on the label (27 CFR 4.62 for wine, ttb.gov, accessed August 2026). Social posts are covered by the same rules, with TTB’s Industry Circular 2022-2 allowing the mandatory information to sit in a readily apparent location such as an account’s About section or a clearly marked link.
Non-alcoholic drinks escape TTB and land under general advertising law instead, where the FTC treats a picture as a claim in the same way it treats a sentence. Show ten visible blueberries beside a bottle whose fruit content is a flavouring, and the image has made a claim the product cannot support.
Platform policy sits on top of all of it. Google’s alcohol policy allows brand and informational advertising for alcoholic drinks only in approved locations and only when it does not target minors, and ads promoting online sale carry further location restrictions (support.google.com, accessed August 2026). None of that is about picture quality, and all of it can pull a campaign that has a perfect hero shot.
The shot list a drink SKU actually needs
Seven frames covers a product page, a marketplace listing and a paid social test. Shoot or generate in this order, because each one constrains the next.
| Frame | Job | Hard constraint |
|---|---|---|
| White-background packshot | Pass marketplace main-image rules | Pure white, no props, no added text |
| Front label panel | Label readable at thumbnail size | Straight-on, no curl distortion |
| Scale reference | Stop the 250ml and 750ml problem | A hand, a glass or a stated size |
| Chilled hero | The category’s conversion shot | Condensation that follows gravity |
| The serve | Show the liquid out of the bottle | Correct colour and head |
| Range or group shot | Sell the other flavours | Consistent height and spacing |
| Social crop | 4:5 and 9:16 versions | Label survives the crop |
The white-background packshot is the one with published rules. Amazon’s main image standard is a pure white background at RGB 255, 255, 255, the product filling at least 85% of the frame, 1,600 pixels on the longest side to enable zoom, and no text, logos, badges, borders or watermarks that are not physically part of the product (sellhound.com, accessed August 2026). Every styling decision that makes a drink photo attractive breaks that rule, so the styled work lives in the secondary images. The full rundown across channels sits in our guide to AI product photos and marketplace rules.
Scale is a category-specific trap. A 250ml can and a 750ml bottle photograph identically when they are alone on white, because nothing in the frame states the size. Returns and complaints follow. One frame with a hand, a glass or a printed dimension solves it.
Condensation is the detail that gives a chilled hero away. Real condensation beads finely near the top of a cold bottle, gathers into larger drops lower down, and runs vertically. Beads of uniform size scattered evenly across the whole bottle read as fake to the eye before anyone can explain why.
Where generative models break on glass and liquid
Glass is the hardest common surface for image models, and the failures repeat in a predictable order.
Label text. Models rewrite lettering. A word becomes a near-word, a percentage moves, kerning drifts. On a drink this is not a cosmetic defect, it is the TTB problem above. Fix by preserving the label rather than generating it: keep the original label pixels and generate the scene around them.
The fill line. Liquid level should sit at one height and stay there across the whole set. Models move it between frames, and a range shot with six different fill heights looks like six different fill volumes.
Refraction. Anything visible behind a bottle should bend and invert through the glass. Models frequently pass the background straight through, so the bottle reads as a flat sticker on the scene.
Cap and closure geometry. Crown caps, screw threads and tamper rings are small repeating structures. They smear first when the model is working at low resolution or the bottle is small in frame.
Condensation and ice. Physically wrong beading, and ice cubes that neither displace liquid nor cast a shadow.
Reflections of a room that does not exist. A generated bottle sometimes carries a reflected window or softbox from the model’s training data. On a product page this is a giveaway.
The workflow that avoids most of this inverts the usual instinct. Do not ask a model to make a bottle. Start from one real photo of your actual bottle, keep the label untouched, and use generation for the parts that are genuinely variable: the surface, the background, the light, the season, the crop. That is the same principle behind AI product photo accuracy, and it is why DesignerBox builds every asset from a customer’s real product photo rather than from a text prompt. Photo Studio handles the angle and scene variants from that one upload, and Spotlight relights an existing frame when the problem is the hotspot rather than the set.
What a bottle image set costs
Drinks sit at the top of the product photography rate card, because food and drink styling is a separate specialism and the shoot needs a stylist as well as a photographer.
| Line item | Traditional range |
|---|---|
| Tabletop or drink photographer, day rate | $1,500 to $5,000 |
| Studio rental with a tethered rig | $500 to $2,000 per day |
| Food or drink stylist | $500 to $1,200 per day |
| Props, surfaces, glassware, ice replicas | $200 to $800 |
| Liquid and label retouching | Per finished frame |
| Flavour extension | A second shoot |
Day rates for styled food and beverage imagery commonly run $1,500 to $5,000, with studio rental as a separate line on top rather than something included in the photographer’s fee (snappr.com, accessed August 2026). Seven frames across six flavours is 42 finished images, which is a multi-day booking before a single retouching hour.
The flavour extension is the line that hurts most and gets budgeted least. Six flavours of one drink share a bottle shape, a cap and a fill height. The only variable is the label. In a studio that is still a full reshoot, because the bottle has to be re-styled, re-lit and re-tethered for every SKU. In a saved workflow it is the same scene re-run against six label uploads. Our product photoshoot cost breakdown runs the same maths across other categories.
On the AI side, cost is credits rather than day rates. DesignerBox plans run Free at 112 credits a month, Basic at $15 for 500, Pro at $35 for 1,000, Premium at $75 for 2,500 and Ultra at $200 for 8,000, and the allocation resets monthly. Video is the expensive operation by a wide margin, since it is priced per second of output rather than per asset, so budget a bottle pour clip separately from a stills set.
Two honest limits before you plan around it. A commercial license starts at the Pro tier, so a Basic plan is for testing rather than for a live listing. Team collaboration, shared brand kits and API access are Ultra features, which matters if a compliance reviewer needs to sit in the same workspace as the designer.
How to run a range of bottles end to end
- Shoot one clean reference per bottle shape. Straight on, evenly lit, label flat to the sensor, at the highest resolution you have. This is the only frame that has to be real, and it is the frame that carries the approved label.
- Photograph each label separately if the shape is shared. Six flavours on one bottle means one bottle reference and six label references.
- Build the white-background packshot first. Get it inside the marketplace rules before anything styled exists, because that image is the one that can block the listing.
- Generate the scenes around the preserved label. Surface, background, light direction, condensation, season. Do not regenerate the label at any point in this step.
- Check the fill line across the set. One height, every frame. This is the fastest visual tell that a range was generated rather than shot.
- Audit against the rulebook that applies. For alcohol: does every depicted label reproduce the approved label, and does the placement carry the mandatory information? For everything else: does any prop in frame make a claim the product cannot support?
- Save the pass as a workflow and re-run it for the next flavour. The point of doing steps one to six carefully is that flavour seven costs one upload.
Saving the sequence is the part that changes the economics. A beverage campaign set produces a six-shot spread from one upload, Beverage Enhance fixes an existing frame that is nearly there, and Styled Scene Generator builds the lifestyle variants once the packshot is locked. Real drink outputs sit in the food and drink gallery, and the category page for drinks brands is DesignerBox for beverage brands. When the same set has to move, Ad Studio cuts it into placement sizes with room for copy.
The pattern generalises to any container-led category. Supplement product photography has the same shape with the FDA in place of TTB, and skincare product photography shares the glass and reflection problems without the regulator. If your issue is a single blown highlight rather than a whole set, start with the product photo lighting fix.
FAQ
How do you photograph a bottle without reflections?
You cannot remove reflections from glass, so you control what it reflects. Place a large diffused light source to the side of the bottle rather than in front of it, which turns into a clean vertical highlight down the shoulder. Add black flags on the opposite side to draw a dark edge. Shoot in a dark room, wear black, and keep the camera far enough back that it does not appear in the curve.
Can you use AI-generated product photos for alcohol advertising?
You can use AI for the scene, but not for the label. TTB requires that a label depicted in an advertisement be a reproduction of an approved label, and it lists label and COLA mismatch among the common issues it finds (27 CFR 4.64, 5.235, 7.235, ttb.gov, accessed August 2026). A workflow that preserves your real label pixels and generates only the background, surface and lighting stays inside that rule. A workflow that lets a model redraw the lettering does not.
What background should a bottle product photo have for Amazon?
Pure white at RGB 255, 255, 255, with the bottle filling at least 85% of the frame and nothing else in shot. No ice, no garnish, no glass, no added text or badges. Amazon also asks for 1,600 pixels on the longest side so the zoom function works (sellhound.com, accessed August 2026). Styled frames go in the secondary image slots where props are allowed.
How many photos does a drink product page need?
Seven covers it: a white packshot, a straight-on label panel, a scale reference, a chilled hero, a serve shot showing the liquid poured, a range or group shot, and a vertical social crop. Fewer than five and the page reads as thin, more than nine and the extra frames rarely change conversion.
Does AI keep the label readable on a bottle?
Only if you stop it from redrawing the label. Text is the weakest area of every current image model, and curved label wrap makes it worse. Upload a real photo of the labelled bottle, keep those pixels intact, and use generation for the room around the product. Check the finished frame at thumbnail size, because a label that is legible at full resolution can still fail at 200 pixels wide.
How much does beverage product photography cost?
Styled food and beverage day rates commonly run $1,500 to $5,000, with studio rental usually charged separately at $500 to $2,000 a day and a drink stylist adding $500 to $1,200 (snappr.com, accessed August 2026). Retouching is billed per finished frame. A six-flavour range at seven frames each is 42 images, which is a multi-day booking rather than a single session.
Do you need a new shoot when a flavour changes?
In a studio, yes. The bottle has to be re-styled, re-lit and re-tethered even when only the label changed. That is why flavour extensions overrun their budget. In a saved workflow, a new flavour is one label upload run against a scene that already passed review, which is the whole argument for building the first set carefully.
Should you disclose that a drink image was generated with AI?
Check the placement rather than the image. Ad platforms set their own disclosure requirements and they change, and some markets add rules on top. The consistent principle across all of them is that the picture must not create a false impression of the product, which is the same standard TTB and the FTC already apply. An image built from your real bottle and your real label is easier to defend on that test than one built from a prompt.
Sources
- TTB, Alcohol Beverage Advertising: What Every Industry Member Should Know, ttb.gov (accessed August 2026). Advertisement definition at 27 CFR 4.61, 5.232, 7.232; prohibited practices at 27 CFR 4.64, 5.235, 7.235; mandatory information at 27 CFR 4.62; social media at Industry Circular 2022-2.
- Google Ads alcohol policy, support.google.com (accessed August 2026).
- Amazon main image requirements, sellhound.com (accessed August 2026).
- Product photography pricing, snappr.com (accessed August 2026).
- DesignerBox plans and credits from the product brief, August 2026.
Regulatory and platform requirements verified from TTB, Google Ads policy and Amazon seller documentation as of August 2026. Rules change, and state law adds requirements on top of the federal ones. Verify against the current source before publishing an alcohol campaign. Individual results vary.
