Render the product shots that must be identical across versions, precise to the drawings, or impossible to film, and film the shots that depend on real hands, real liquids and real light. In most video ads that means a rendered hero reveal, cutaway, colorway lineup and packshot, plus a filmed tabletop or lifestyle shot for pours, textures and use. Decide each shot on five things: the inputs you have, the variants you need, the motion, the schedule, and whether the shot is offered as proof.
Most “3D rendering vs product photography” pages compare catalog stills. A video ad changes the math: a 4-second shot at 24 frames per second is 96 frames, and every frame of a render has to be computed. This article covers the comparison for motion, the truthfulness rules that apply to a rendered product, and the open formats that let one model serve the ad, the product page and AR. Rules and documentation are as of November 2026.
The short answer: five factors decide each product shot
Neither method wins for a whole spot; each wins for certain shots. Score each product shot on inputs, variants, motion, schedule and proof, then pick the method that fits, which usually gives you a mix.
| Factor | Filming on a tabletop or set | Rendering from a 3D model |
|---|---|---|
| Inputs | A finished, camera-ready sample in its final finish and packaging | CAD or a scan, plus material, finish and label references |
| Variants | Each colorway or pack must exist physically and be shot or retouched | Swap materials or labels in the same scene and re-render |
| Motion | Real pours, condensation, fabric, food and hands | Impossible camera paths, exploded views, cutaways, transitions between variants |
| Schedule | Compressed into prep and shoot days, but waits for the sample | Spread across modeling, look development and render hours, but can start from drawings |
| Proof | Viewers read it as the real thing, so a demonstration is straightforward | Reads as the product when photoreal, so it must match the product exactly |
Two questions settle most cases before you get to the table. Does the product exist in its final form yet? If not, render. Is the shot offered as evidence that the product does something? If so, film it as it happens, or have counsel review the render.
The rest of this article takes each factor in turn and ends with a shot-by-shot plan and a checklist. The general choice between filming, animating and rendering a whole spot is covered in animation, live action or both.
What each method needs before the first frame
Filming needs a finished product you can light; rendering needs accurate data and references you can match. Which of those you have, and when, often decides the method before any creative preference does.
Filming needs a finished, camera-ready sample
A tabletop shoot photographs exactly what arrives on set. The sample has to be the final color, finish, label and packaging, and usually needs several units: one hero unit kept clean for close-ups, others for pours, drops or anything that marks the surface. Dust, fingerprints, scratches and molding marks show at macro distance, so samples are cleaned and prepared before they are lit.
Commercial estimates have lines for this work. The AICP Bid Form lists “Product Prep / Color Correct”, prop fabrication and home economics supplies among its expense lines, and the AICP Bidding Reference Guide (February 2026) asks at the first call whether any product specialists are needed, giving “pour specialist” as its example. If a product needs a specialist to look right on camera, that is a cost and a scheduling dependency.
The biggest dependency is time. If samples arrive from the factory a week before launch, the shoot waits for them, and so does every shot that depends on the shoot.
Rendering needs data and references
A render is only as accurate as what it is built from. The minimum set is:
- Geometry: CAD from the product’s engineering files, or a 3D scan of a physical unit. It should be the released design, not an early prototype.
- Material and finish references: a physical sample or finish swatches, the approved color specifications, and notes on gloss, texture and transparency for each part.
- Graphics: label and packaging artwork as flat files, with dielines where the label wraps a curved surface.
- Reference photographs: the real product under neutral light from several angles, so the render can be compared side by side.
- Visibility notes: which parts the camera will see, so modeling time goes where the shots need it.
Without a physical reference, a render matches the drawing, not the product. That is fine before samples exist, but plan a comparison step once the first production unit arrives. Converting engineering CAD into a clean model for animation is its own process, with its own pitfalls, covered in CAD to product animation.
| Input | Needed to film | Needed to render | Usually provided by |
|---|---|---|---|
| Final production sample | Yes, several units | Strongly recommended as reference | Brand or manufacturer |
| CAD or scan | No | Yes | Engineering, or a scanning step |
| Color and finish specifications | Helpful for color correction | Yes | Brand or product design |
| Label and pack artwork files | Only if labels are made for the shoot | Yes | Brand design team |
| Claims and their evidence | Yes, for any demonstration | Yes, for any demonstration | Brand, regulatory or legal |
| Placement list and aspect ratios | Yes, for framing | Yes, for camera setup | Marketing or media team |
Variants: colorways, packaging and regional versions
Variants favor rendering. A colorway or a regional label becomes a material or texture swap in the same 3D scene, while filming needs every physical variant on set and either a full setup per variant or frame-by-frame retouching.
How variants work in a 3D scene
In 3D, the geometry and camera stay fixed and the surface changes. The open formats now encode this directly. The Khronos Group lists a Material Variants extension among the projects of its 3D Commerce Working Group, so one glTF file can carry several material sets. OpenUSD describes VariantSets, which “allow an asset creator to bundle different multiple variations of an asset all into a single package with a ‘variant selector’” that downstream users can switch.
Variants are not free in 3D. Every colorway rendered in motion adds its own frames, and every one needs a color check against the approved reference. What disappears is the reshoot: no new samples, no new setup, and the camera move is identical across colors because it is the same virtual camera. Plan one color approval per variant: each colorway’s render is compared with its own physical swatch or sample, signed off by the person who owns that color.
How variants work on a tabletop
On a tabletop, each colorway is a separate physical object. A repeatable camera move needs a motion-control rig that can run the same path once per unit, or a locked-off camera. The alternative is to shoot one color and recolor the others in post, which works for flat matte surfaces and gets harder with gloss, metal and transparency, where reflections carry the color too.
What the render arithmetic looks like
Example (fictional). Wrenfield is an invented kitchenware brand launching a kettle in 5 colorways; all figures are illustrations. The spot’s product moments total 6 seconds: a 3-second reveal and a 3-second spout close-up. The example assumes 8 minutes per frame on one machine and 4 machines rendering at once.
- Every colorway in motion: 6 seconds x 24 fps = 144 frames per colorway; 144 x 5 = 720 frames. 720 x 8 minutes = 5,760 machine minutes = 96 machine hours, or 96 / 4 = 24 clock hours per full pass.
- Lead color in motion, plus a lineup: 144 frames for the lead color, plus a 3-second lineup of all five kettles (3 x 24 = 72 frames) = 216 frames. 216 x 8 = 1,728 machine minutes = 28.8 machine hours, or 7.2 clock hours per pass.
- Lead color only: 144 x 8 = 1,152 machine minutes = 19.2 machine hours, or 4.8 clock hours per pass.
The middle option shows every color for a fraction of the render time. The choice is a media decision as much as a production one: if each colorway runs as its own ad, it needs its own motion shots; if the colors share one spot, a lineup may be enough.
Packaging and regional versions
Packaging changes follow the same logic. A new language on a label is a new texture on the same model, which is why rendered packshots are common in campaigns that run in several markets. Filmed packshots need a printed sample of each version, and small text on a curved label is hard to replace convincingly in post. The versioning matrix calculator shows how fast versions multiply once languages and ratios are added.
Motion: what a tabletop camera can and cannot do
Filming wins when the shot depends on physical behavior: real pours, condensation, foam, fabric drape, food texture and hands. Rendering wins when the camera has to do something physics will not allow: pass through the product, separate its parts, see inside it or morph one variant into another.
What filming does well
A camera records how liquids, steam and fabric really behave, and audiences are good at spotting the difference. A pour that splashes the way coffee splashes, a drop of condensation running down a cold can, or a knit fabric stretching under a thumb all read as true when filmed. High frame rates turn a fraction of a second into a slow, readable moment, and a motion-control rig can repeat a camera path for several takes or passes.
What rendering does well
A virtual camera has no size, no weight and no lens limits. It can glide through a gap narrower than any real lens, orbit a product in one continuous move, or hold perfect focus across a macro shot. Exploded views, cutaways, x-ray views and transitions between colorways are native to 3D and hard or impossible on a tabletop.
3D can simulate liquids, cloth and particles too. Those simulations are separate setups that must be tuned, cached and reviewed before rendering, and every round of tuning adds time before the first final frame exists. If the shot’s whole point is a convincing pour, filming is usually the shorter path.
Hybrid shots: real liquid, rendered product
A common middle path films the physical behavior and renders the product. A real pour is shot into a stand-in vessel and the rendered product is composited over it, or a real hand with tracking markers holds a stand-in that is replaced by the render. The liquid and the hand stay real and the product stays exact, at the cost of tracking and compositing on every frame.
The proof question still applies. If the pour demonstrates a feature, such as a spout that does not drip, the filmed pour has to come from a real product that performs that way, not from a stand-in that behaves better.
| Move or effect | Filming | Rendering | Note |
|---|---|---|---|
| Pour, splash, steam, condensation | Strong | Possible with simulation | Simulation adds setup and review time |
| Fabric stretch, drape, texture under a hand | Strong | Possible with cloth simulation | Real hands are hard to render |
| Orbit or turntable of the product | Good with a motion-control rig | Strong | Rendered orbit is identical for every colorway |
| Camera through a gap or inside the product | Not possible | Strong | Shows internal structure |
| Exploded view or cutaway | Needs a sectioned prop | Strong | Must match the real construction |
| Transition from one colorway to another | Needs a cut or a dissolve | Strong | A material change can animate over time |
| Macro on a tiny part with deep focus | Limited by lens and depth of field | Strong | Focus is a render setting |
| Product in use outdoors | Strong | Possible as a composite | A real place keeps real light |
Schedule: render hours against shoot days
Filming compresses image-making into prep and shoot days but cannot start until samples exist. Rendering spreads the work over modeling, look development, animation and render hours, and it can start from drawings weeks before the first production unit.
How render time is counted
Render time is arithmetic, and it is worth doing before you approve the plan:
- Frames = seconds of 3D x frames per second.
- Machine time per pass = frames x minutes per frame.
- Clock time per pass = machine time / machines rendering at once.
- Total = clock time per pass x the number of full passes.
Example (fictional). The Orlenna Ridge Bottle spot from the site’s example campaign has 11 seconds of 3D at 24 fps = 264 frames. At 6 minutes per frame, one pass takes 264 x 6 = 1,584 machine minutes, or 26.4 machine hours. On 4 machines that is 6.6 clock hours per pass. Three full passes take 19.8 clock hours and 79.2 machine hours. Orlenna is invented, and these are illustrations, not quotes.
The render time calculator runs the same steps with your numbers. Its note is worth repeating: the result assumes frames split evenly and none fail, so add time for re-renders after notes and for compositing passes rendered separately.
Why passes matter more than minutes
A pass is one full render of every frame. Plans usually include at least an early pass for review, a pass after notes and a final pass, and most of the schedule risk sits between them: a note on a material after the final pass means rendering again. Reviewing untextured grey renders for camera and timing, then a few still frames for materials, before committing to full passes keeps late notes cheap.
A side-by-side schedule
How each method is estimated
Filming is estimated in days and in the people, space and equipment for each day; rendering is estimated as a one-time model build plus a cost per shot and per variant. Which comes out lower depends on how many product shots and variants the campaign needs, and on whether the model will be reused.
The lines in a filmed estimate
A tabletop or studio shoot uses the same sections as any commercial on the AICP Bid Form: prep and shoot crew, studio costs (with separate lines for build, pre-light, shoot and strike days), equipment rental, art department labor and expenses for surfaces and sets, and props, where product prep sits. Post adds cleanup of dust, seams and reflections, and color correction to the reference.
Most of those lines are day rates. A well-planned tabletop day can cover several setups, so the cost per shot falls as more shots fit into the day, and a second day is a step up. A motion-control rig, a specialist or a high-speed camera adds its own line for every day it is on set.
The lines in a rendered estimate
A rendered estimate usually separates five kinds of work. The model build happens once per product, scaled to how close the camera gets. Look development sets up materials and lighting, once per product and again for each variant. Each shot then needs layout, animation and lighting, the render itself costs machine hours, and compositing joins the passes.
The model and its materials carry over to every later shot and campaign if they are handed over and licensed for reuse; the per-shot lines do not. That is the economic core of the rendering case: the more shots, variants and campaigns share one model, the smaller its share of each.
Example (fictional). The site’s production budget worksheet uses illustration values, not prices or market rates. Its 3D lines are a 3D model from product drawings at $2,400 flat, plus 3 3D shots x $3,200 = $9,600, for $12,000 in total. Its live-action lines are 1 day of crew at $9,500, 1 day of camera, lighting and grip at $2,800, and $1,200 for locations, permits and insurance: $13,500.
With that structure, a fourth rendered shot, say a packshot in a second color, adds 1 x $3,200 = $3,200 and leaves the model line unchanged. A later campaign that reuses the handed-over model starts without the $2,400 line. A later campaign that films the product again starts with another day: $9,500 + $2,800 = $12,300 before locations.
Questions that make bids comparable
- How many hours of model build are assumed, and to what level of detail?
- What minutes per frame, how many machines and how many full render passes?
- How many material review rounds, and on what kind of frames?
- For filming: how many setups per day, how many samples, and is retouching included?
- Which files are handed over at the end, and may they be reused in other media?
Common failure points and how to check for them
Renders tend to fail by looking too perfect, or slightly wrong in color, scale or label; filmed product shots fail on sample defects, reflections and color shifts under studio light. Each method has review checks that catch most problems before the final pass or the wrap.
Where renders go wrong
- Surfaces too clean. Real products carry tiny scratches, dust and edge wear. Without any, a render reads as computer-made. Subtle imperfection, signed off against the real sample, fixes it.
- Color drift. The same material reads differently under different lighting setups and on uncalibrated screens. Approve color on a still frame under neutral light, next to a photo of the production sample. Khronos says its PBR Neutral Tone Mapper was designed for this kind of true-to-life product color.
- Label distortion. Artwork wrapped onto a curved surface stretches, and small type blurs at delivery resolution. Build labels from the dielines and check every line of type at the final size.
- Wrong scale or physics. A product slightly too large in a hand, or a liquid moving too slowly, makes the shot feel wrong even when viewers cannot say why.
- The wrong revision. A model built from a pre-release drawing goes out of date when a button moves or a seam changes. Record the drawing version in the plan and check it again before the final pass.
Where filmed product shots go wrong
- Sample defects at macro distance. Molding lines, scratches and misprinted labels show in close-ups. Budget product prep and post cleanup, and bring spare units.
- Reflections. Glossy and metallic products reflect the studio, the crew and the camera. Flags, cards and post cleanup handle it, at a cost in time.
- Color shift. Lights and camera processing change how a color records. Keep the reference sample on set and have the colorist match to it.
- Unit-to-unit differences. Two samples of the same product can differ slightly, and a cut between them shows it. Use one hero unit for matching shots.
- Physical limits. A pour does not land on its mark every take, and a product will not balance where the board shows it. Allow time for takes.
Review steps that catch most problems
- Grey review: untextured renders, or a rehearsal with a stand-in, to sign off camera, framing and timing.
- Material review: still frames compared with the production sample, on a calibrated display, by the person who owns the product’s color.
- Type and label check: every label and legal line read at delivery resolution.
- Full-resolution review: the hero seconds watched frame by frame before the final pass, or the selects reviewed before the set is struck.
Truthfulness: a render has to show the product as it is
US advertising law judges what the viewer takes away from the whole ad, pictures included, not how the picture was made. A render must not make the product look better than it is or show it doing something it does not do, and a render offered as proof is a demonstration.
The deception standard applies to imagery
The FTC’s Policy Statement on Deception (October 14, 1983) sets three elements: a representation, omission or practice likely to mislead, judged from the view of a consumer acting reasonably, that is material. It presumes express claims are material and says information about “durability, performance, warranties or quality” is likely to be material. It also quotes a court of appeals on the Commission’s right to scrutinize “the visual and aural imagery of advertisements”, because ads are judged as a whole.
For product renders, the material facts are often visual: the true color, the finish, the size relative to a hand, what comes in the box. A render that is accurate to the CAD but shows a finish the factory does not produce, or an accessory sold separately, can mislead as surely as a false sentence.
Mock-ups and the Colgate line
In FTC v. Colgate-Palmolive Co., decided April 5, 1965, a shaving-cream commercial appeared to shave sandpaper. The prop was sand on plexiglass, and real sandpaper needed about 80 minutes of soaking. The Supreme Court upheld the Commission’s view that the undisclosed mock-up was a separate misrepresentation, because viewers were told they were seeing proof.
The Court also drew a useful boundary. Actors eating mashed potatoes that look like ice cream are not covered, because the prop is not offered as proof of a claim. But when a commercial “invites the viewer to rely on his own perception, for demonstrative proof of the claim”, undisclosed props “in strategic places might be a material deception”. A render is not a prop, but the same reasoning is the safest way to read it, an inference your counsel should confirm: a render that illustrates is one thing, a render offered as a test result is another.
Claims need evidence before the ad runs
The FTC’s Policy Statement Regarding Advertising Substantiation (November 23, 1984) reaffirms that advertisers and agencies must have a reasonable basis for objective claims before they are disseminated. When an ad says “tests prove” or similar, the advertiser needs at least the level of support the ad describes. A rendered animation of water beading off a coating, or of heat staying inside a wall, makes a performance claim whether or not a word is spoken.
Why a small “simulated” super is not a fix
A disclaimer in small type does not rescue a misleading picture. The same FTC statement says written disclosures or fine print “may be insufficient to correct a misleading representation”, that “pro forma statements or disclaimers may not cure otherwise deceptive messages”, and that qualifying disclosures “must be legible and understandable”. If a shot only works with a disclaimer, the safer production choice is usually a different shot, and whether a disclosure is needed at all is counsel’s call.
Practical rules for rendered product shots
These are production habits, not legal advice; your counsel decides what a given ad needs.
- Build from the released design. Use production CAD or a scan of a production unit, and record which version of the drawings the model came from.
- Match the production finish. Compare renders with a production sample under neutral light, and keep the signed-off comparison frame.
- Show only what ships. Leave out accessories, colors and features that are not part of the advertised product, or make clear they are sold separately.
- Keep scale honest. Hands, tables and environments around the product should be at true size.
- Treat rendered performance as a claim. If a render shows the product doing something, send the shot and the evidence behind it to the people who review claims.
- Decide labels per destination. Platform AI-disclosure rules target realistic generated or heavily altered content; if generative tools made any part of a realistic product shot, check each platform’s current rule and record the decision.
One model, many uses: glTF, OpenUSD and the handover
A product model built for an ad can also serve the product page, AR viewers and the next campaign, if it is delivered in open formats with its materials documented. That reuse is a real part of the rendering case, and it has to be planned at the start, not requested at the end.
glTF for web, ecommerce and Android AR
The Khronos Group describes glTF as “a royalty-free specification for the efficient transmission and loading of 3D scenes and models by engines and applications”. A scene can be stored as a single binary .glb file, and glTF 2.0 was published as the ISO/IEC 12113:2022 International Standard. Khronos also publishes a PBR Neutral Tone Mapper, which it says is designed to display physically based renders with true-to-life colors for ecommerce, architecture and CAD.
Google’s Scene Viewer, which shows 3D models in AR on Android, takes a URL to “the glTF or glb file that should be loaded”. A glTF export of the ad model is therefore the usual starting point for a product page viewer.
OpenUSD for production and Apple AR
OpenUSD is the open-source Universal Scene Description project. Its introduction describes it as software to “interchange and augment arbitrary 3D scenes” made of many assets, with composition features such as references and VariantSets. Apple’s AR Quick Look page says built-in apps such as Safari, Messages, Mail, News and Notes use Quick Look to display USDZ files in 3D on iPhone, iPad and Apple Vision Pro.
What to ask for at handover
A model built for film-quality renders is usually heavier than a real-time viewer can load, so plan an optimized version if you want ecommerce and AR use. Agree the list in the contract, then expect:
- The source scene files, as the contract defines them.
- A glTF or GLB export and a USDZ export, if real-time use is planned.
- Texture maps and a material sheet: each part’s color, finish and gloss values.
- The variant list, with the reference each color was matched to.
- A license list for any purchased models, textures or HDR images used in the scene.
- The render settings for the hero shots, so a later re-render matches.
What transfers, and on which terms, belongs in the contract and the rights sheet. The deliverables page lists what a handover usually includes, and rights and usage covers ownership of 3D assets.
Deciding shot by shot: the hybrid most product ads use
Most product ads end up mixed: rendered hero, detail, variant and packshot moments, with filmed hands, pours and lifestyle scenes. Write the decision per shot, with the reason, so bidders price the same plan.
| Shot | Usual choice | Reason |
|---|---|---|
| Hero reveal | Render | Identical in every version, perfect light, no sample needed |
| Detail or macro of a feature | Render, or film if the texture is the point | Deep focus and exact geometry, unless the material must read as real |
| Pour, splash or food | Film | Real physical behavior; specialists where needed |
| Hands opening or using the product | Film, or composite if the product is not ready | Real hands read as real |
| Colorway lineup | Render | One scene, one camera, every color |
| Exploded view or cutaway | Render | No camera can see inside |
| Product in a real place | Film, or composite | Real light and weather |
| Demonstration offered as proof | Film as it happens | Viewers rely on what they see |
| Packshot and end card | Render | Re-renders for each language, ratio and color |
Example (fictional). The site’s example campaign plans an invented 30-second spot for the Orlenna Ridge Bottle. Shot 4 (12 to 17 seconds) is a 3D exploded view built from the product drawings: outer steel, vacuum gap, copper lining, inner steel, because a camera cannot show a vacuum gap. Shot 8 (27 to 30 seconds) is a 3D packshot, identical in every version and re-renderable for a new color.
The proof stays filmed. Shot 6 (20 to 24 seconds) is the summit pour: the same coffee in the same bottle, filmed hours later, with the steam still there. Shot 5 (17 to 20 seconds) is the blend: a real hand holding the 3D bottle, because the hand must look real and the bottle must match shots 4 and 8.
That split is typical of the reasoning, not a rule. A brand whose product is a fabric or a food may film nearly every product shot; a brand launching hardware before production may render nearly all of them and add filmed hands later. Mixed-media production explains how the blended shots are planned.
Checklist before you request bids
Settle the inputs and the shot list before you ask for prices, because both methods are priced from them. A bidder who has to guess whether samples or CAD exist will price the risk.
- Confirm what exists. Production samples (how many, when), released CAD, color and finish specifications, label artwork.
- List every product shot with its seconds. Seconds drive render time and shoot time alike.
- Mark the method per shot using the five factors, and write one sentence of reason.
- Count the variants that need motion. Colorways, packs and languages; decide which run as separate ads and which share a lineup.
- Do the render arithmetic. Frames, minutes per frame, machines, passes; ask bidders to state their assumptions.
- Identify every demonstration. Note the claim, the evidence and who reviews it.
- List the reuse you want. glTF, GLB or USDZ exports, material sheets and license lists, in the contract from the start.
- Name the approvers. Who signs off the grey review, the material match and the final color.
The consumer products page covers launch work where samples arrive late, and the service pages for 3D and CGI and live-action production describe each side of the split. When the list is ready, you can send a brief with the samples and files you have, and the plan that comes back assigns each product shot to a method with its reason.
