Start with what a buyer needs to see
A launch rarely needs just “better pictures.” A buyer may need to identify the exact variant, understand the included pieces, judge a finish, or imagine the item in a room. Those jobs require different evidence. A generated living room can provide context; it cannot verify the rear connector of a speaker photographed only from the front. A studio photograph may supply that missing evidence without producing every campaign composition.
Write the destination and buying question beside each proposed image. For a marketplace main image, inspect the current category requirements before choosing a visual treatment. For a supporting banner, decide whether the complete product and an approved message can fit the intended crop. Adobe's product-photography guide distinguishes product-only photographs from lifestyle work; treat that distinction as a planning starting point, not a fixed rule for every channel.
Classify the source set before choosing a service
| Source condition | Practical route | What still needs checking |
|---|---|---|
| Sharp current product, complete required view | Try a restrained background or scene treatment | Product changes, edges and delivered crop |
| Correct item but unreadable identifying label | Capture a readable label/front view | Glare, variant wording and relationship to the whole item |
| Requested back or open state is missing | Targeted photography of that state | Exact revision, assembly and visible construction |
| Required picture must retain captured markings | Original-photo masking/compositing or direct photography | Mask quality, shadows and export |
| Untested visual direction only | Concept generation, clearly kept separate from product evidence | Do not release a plausible prototype as the shipped item |
Separate “unknown” from “unattractive.” A cluttered background around a well-recorded item is a presentation problem. A hidden hinge is an information problem. Changing the background can address the first without establishing the second. More detailed prompts do not convert a guess into evidence.
Worked route decision: a desk-lamp launch
This is an illustrative planning exercise, not a completed shoot or AI test. A seller has a sharp front photograph of the current lamp, a blurry rear photograph and an old lifestyle image showing a discontinued switch. The launch needs a product identification image, a rear-port detail and two mood banners. The attractive old scene is excluded from product evidence because the switch revision differs.
The seller assigns the identification image to the real front photograph, using source-preserving cleanup if needed. The rear detail stays blocked until a new photograph shows the current connector, cable exit and adjoining panel. The two mood banners may use AI product photography based on the current front view, with no rear rotation or demonstration of an unverified dimming function. Their briefs reserve separate copy space and keep decorative electronics out of the offered contents.
Approval is per deliverable. If a generated banner alters the switch, it fails even when its room looks excellent. If the supplier's new rear photograph belongs to old stock, it fails the evidence request. The route decision therefore creates a short capture request and two bounded scene briefs, not a blanket statement that either technology is better.
Compare the cost of accepted deliverables
Make two estimates using your actual quotes. A shoot estimate includes product preparation, shipping, capture, selection, retouching and rights needed for the use. An AI estimate includes source preparation, generation, rejected candidates, factual review, external repairs and layout finishing. Do not compare one raw generation with a studio package that includes finished approved assets. Do not assume a subscription removes review labor.
For each route, write the stop condition before paying for another attempt. An example is “one pilot scene must retain the source-visible switch and readable variant name.” If it fails repeatedly, simplify the brief or retain the original product photograph in a suitable editor. Choose a specific capture request when the blocker is missing evidence. Keep quotations and time assumptions dated; the method does not assert a universal saving or current CartVisual price.
Run a small, reviewable pilot
- Select one representative difficult item, not only the easiest matte box.
- Record its current SKU revision, authoritative views and required image role.
- Use the public creative brief planner to separate facts from scene choices. It drafts in the browser; it does not inspect the image or generate media.
- Open CartVisual Product images for the selected scene workflow. Review account requirements and the displayed quote before generation.
- Compare the exported result with the source and the destination preview. Record factual failures separately from aesthetic preferences.
A useful pilot answers whether this route can produce the specified asset with acceptable review work. It does not establish an accuracy rate for every material or SKU. Glass, reflective surfaces, fine fibers and small printed markings may need different handling. For exact captured product pixels, use an appropriate source-preserving process and inspect the cutout rather than assuming scene generation leaves the object untouched.
Close the decision with an asset list
End with a list stating which files are approved, which need photography and which remain concepts. Attach the source revision and permitted placement to each approved file. A front-only mood image should not later become a rear-detail demonstration simply because it has a high-resolution export. Retain original captures so a new offer or packaging revision can be checked against real evidence.
If the launch cannot wait for an unavailable angle, remove that unsupported view and use a supported composition. This is a scope choice, not a promise that customers no longer need the missing information. Document the limitation and decide separately whether the listing can responsibly answer the buyer's question before it goes live.
Continue with the next decision
Sources and scope
The worked decisions are illustrative editorial exercises, not generated-image tests or measured campaign results. The practical checks in this guide are the seller’s responsibility.
- Adobe: product photography: Product-only and lifestyle photographs have different purposes; plan source capture and lighting.
- Photoroom: making AI product images realistic: Clear source imagery and separate realism from product fidelity; no vendor performance figures adopted.