Consider a hypothetical launch sourced from a Chinese apparel supplier: a printed top, a tailored jacket and commuter trousers. A brief asks for the same three model clips for every style. Meanwhile, a separate team orders AI outfit variations. Both teams can deliver exactly what was requested while leaving the shopper's most important question unanswered.
The useful budgeting unit is not simply a photograph, a clip or a SKU. It is the purchase question the content needs to resolve. ONCE proposes a three-question allocation method to decide where to use styling previews, where to repair product information and where physical samples deserve camera time. This is a proposed planning method, not a report of a client launch or measured savings.
Personalization changes the preview, not the meaning of fit
On December 11, 2025, Google introduced a selfie-based digital body option for US shoppers using its virtual try-on experience. A personalized image can make an outfit easier to imagine. That announcement does not establish that a shopper's size or comfort has been verified.
Google's current merchant guidance for uploaded-photo Try-on says it does not establish fit or suggest size. It describes non-sponsored product results, not Shopping Ads. Separately, apparel try-on using Google's photographed models can appear in qualifying Shopping ads. These are different experiences. The help pages were checked on October 18, 2026; neither displays a publication date.
Do not turn either description into a claim that all try-on features work in paid placements. Nor is a shopper experience a ready-made campaign asset library: the uploaded-photo merchant guidance says merchants cannot access users' images or generated looks. A brand producing its own advertising needs separately authorized inputs and its own production arrangements.
Our inference is narrower than “AI replaces fashion shoots.” As more styling previews become available, the remaining physical shoot should be commissioned for a more specific job.
Route three questions to three kinds of work
Ask merchandising, customer service and creative to state the unresolved question in a shopper's words before choosing the medium.
| Purchase question | First production response | What it must not quietly imply |
|---|---|---|
| Does this work with my style or wardrobe? | Product-checked styling previews and outfit exploration | That a convincing image verifies a particular size |
| Which size should I order? | Approved garment measurements, a defined measuring method and fit notes; any human reference needs relevant measurements and worn size | That an AI body image is a sizing recommendation |
| What happens when I sit, reach or walk? | A physical-sample sequence designed around the relevant action | That generated movement verifies stretch, ease or comfort |
This is a routing rule for the information available on the project, not a prediction of permanent technical limitations. Physical footage also has limits. It shows a particular sample, wearer and action. It does not guarantee fit for every body, and visible movement alone cannot prove subjective comfort.
For overseas teams, make the second column part of the supplier handover. “Send the medium sample” is not enough. Establish which garment version the sample represents, what measurements have been approved and which claim the footage is meant to illustrate. A production team should not invent fit guidance from appearance.
Give sample time to the highest-priority unanswered decision
Return to the hypothetical three-style launch. The printed top might need outfit exploration once print placement and product details are checked. But if transparency is the unresolved concern, arrange a real demonstration with lighting and underlayer conditions made clear. A loose silhouette does not answer that question.
For the jacket, a workwear promise may justify filming a seated reach while the jacket is fastened. For the trousers, clarify whether the question concerns length, the waistband when seated or the outline after putting an everyday object in a pocket. These are different shot requests, not interchangeable “lifestyle movement.” Avoid extending what the visible action shows into a broad performance claim.
Prioritize with three prompts:
- Would the answer change the buying decision?
- Is that answer already present in reliable product information?
- What new visible information would this shot add?
If the missing answer is a measurement, ask the product team for the measurement. Shooting more variations is not the remedy. If actual support questions or return-reason records exist, use them to inform the order. Without such records, label the priority as a hypothesis rather than presenting it as a known customer pattern.
Reuse by construction, not by a shared style name
A single physical sample may seem sufficient for every colorway. Before reusing its footage, have the product owner compare pattern, fabric composition and weight, lining and relevant construction. Even matching specifications do not automatically establish identical opacity or surface appearance across colors.
A new lining, for example, can make the old drape reference inappropriate. Reuse only the information confirmed to remain applicable; separately capture differences that could change the purchase decision. Generated variants still need a product-detail check so that a preview does not add a pocket, alter a seam or remove trim.
Plan clear source images separately from editorial campaign images. Google's current merchant guidance asks for the full garment and unobscured details. A dramatic crop may serve a campaign without serving that input requirement.
The human reference also deserves a deliberate casting brief. ONCE's model casting information is a starting point for discussing talent and usage arrangements—not proof that any selected model represents the fit needs of all customers. Select the wearer and action for the specific question being filmed.
Buy a resolved information task, not a larger file count
Separate the proposal into source-image preparation, generated styling and review, approved product-information preparation, physical-sample capture and channel versions. Define the output and revision work for each. This makes the trade-off inspectable without inventing a universal savings percentage.
For a production partner, a longer-term service worth testing is maintaining the relationship between purchase questions, garment versions and usable content. Whether that supports recurring revenue or better margins depends on demand, maintenance effort and willingness to pay. A platform announcement cannot prove that business outcome.
Before the next collection leaves the supplier, send ONCE the style list, approved product information and one unresolved purchase question for each priority item. Then choose which questions need previews, which need better data and which justify shipping a physical sample. The goal is not to make the real shoot compete with AI on image volume. It is to use each kind of work where it actually helps the shopper decide.
