AI product photography can help an agency explore settings and campaign compositions around a client's product. The difficult part is preserving the facts: the pack's shape, label, material, and included components must remain recognizable and accurate.
Start with approved source images and a clear role for each output. Then choose whether to generate a complete scene or combine an original product asset with a generated environment. This workflow explains both routes and gives you prompts, review criteria, and a handoff record to adapt.
The worked brief below is fictional. It describes a process and suggested prompts, not measured generation results or evidence of a completed client campaign.
Gather the source material first
Ask the client for a clean product photo, any alternate views needed, brand assets, approved product descriptions, and the required placements. Record which asset is authoritative if the packaging has changed. A previous campaign image may show a discontinued label or an old offer.
Keep the brief beside those files. Specify what the image should communicate, where the product should sit, and which details cannot change. Download the product photography worksheet to record the inputs, prompt, checks, and final handoff.
| Input | What to establish |
|---|---|
| Product reference | Approved pack, proportions, colors, material, and visible details |
| Image role | Clean hero, use context, detail, or campaign composition |
| Setting | Surface, background, props, and the intended visual direction |
| Message | Approved facts and separately editable copy |
| Placement | Required dimensions, crop, file type, and destination rules |
| Review | Accuracy reviewer, design reviewer, and final approver |
If an ingredient, claim, or measurement is missing, leave it unresolved. An image-generation prompt is not a source of product facts.
Choose between generation and compositing
Reference-guided scene generation is useful for exploring a visual direction. Supply the photo to a model that supports references, describe the scene, and inspect the output against the original. Reference support is model-dependent; an uploaded image does not imply exact preservation.
Compositing with the original product is useful when packaging must remain unchanged. Keep the approved product asset, remove its background if needed, generate or choose a setting, and arrange the original asset in the composition. Check edges, perspective, scale, light, and contact shadows so the product belongs in the scene.
The second route avoids accepting an invented label, but it still requires design judgment. A product cutout placed on a new background can look implausible if the lighting and viewpoint disagree.
In Quickmake, use image generation for the scene and background removal when preparing the supplied asset. Keep the original accessible for comparison.
A fictional coffee campaign brief
Suppose a fictional brand, Soleil Café, needs three directions for a coffee pack: a clean hero, a café setting, and a banner composition. The client supplies an approved pack photo and logo. The exercise makes no claims about roast, origin, ingredients, or product performance.
| Direction | Visual job | Constraint |
|---|---|---|
| Studio hero | Make the pack easy to recognize | Simple surface and background; no invented callouts |
| Café setting | Suggest a moment of use | Props support the scene without implying included accessories |
| Banner composition | Leave room for the campaign message | Add approved copy separately after image selection |
Keep these as distinct composition questions. A clean hero and a lifestyle concept should not be judged by the same amount of decorative detail.
Write a prompt around the visual task
For a reference-guided studio concept, adapt this prompt to the selected model:
Use the supplied product photo as the reference for the coffee pack. Create a clean studio product concept on a warm cream surface, with soft light from the upper left and a quiet background. Keep the pack prominent. Do not add promotional badges, new packaging copy, accessories, or product claims. Preserve the visible shape and label as closely as possible. The result will be compared with the original before use.
Preservation wording expresses the requirement; it does not guarantee the result. Inspect the image at a size where the label and edges can be read.
For the compositing route, request the environment without asking the model to reproduce the pack:
Create a simple cream-colored tabletop and warm neutral backdrop for a product composition. Use soft light from the upper left. Leave an uncluttered central area for an upright coffee pack and additional space on the right for a short headline. Include no product, text, logo, badge, or promotional claim.
Place the approved product asset into the resulting scene. Adjust the composition around the product rather than stretching the pack to fill an unsuitable space.
Plan a product listing set in Quickmake
Quickmake includes a Product Listing Images workflow. Supply the product, the marketplace or destination, approved reference images, and the purpose of each deliverable. A useful sequence may include a clean hero, a context image, and a detail composition.
The workflow provides a planning structure; every output still needs review. Supply destination requirements instead of assuming one set of rules applies to every marketplace. When the destination prohibits particular backgrounds, props, or overlays, keep those constraints in the brief.
For a single campaign visual, you can also generate a scene and compose it directly. Use the design editor to add separate headlines, shapes, and logos. Separately added text stays editable; text rendered inside a generated image does not automatically become a text object.
Review fidelity before reviewing style
Check these areas against the approved photo:
- Shape and proportions: Has the pack become wider, taller, or a different construction?
- Packaging: Are the label, closure, logo, and visible text consistent with the source?
- Materials and color: Does the image represent the actual finish rather than a visually appealing substitute?
- Components: Did generation add handles, accessories, ingredients, or items the product does not include?
- Scene logic: Do the perspective, light, scale, and shadows make sense?
If the product is inaccurate, revise or switch to the original-asset composition. Do not approve the image because its setting looks polished. Record the failed detail so the next attempt has a specific correction goal.
Once fidelity is acceptable, evaluate hierarchy, negative space, and how the image supports the message. Review the final crop, not only the full scene shown on the canvas.
Keep revision and handoff understandable
Name each direction, record its source assets, and retain the prompt and model used. A client comment should identify the version and the detail to change. “Pack label differs from the approved asset” is more actionable than “looks wrong.”
Before export, verify dimensions, text readability, logos, and destination requirements. Keep the original pack and the composition available if further edits are expected. A flat export is a delivery file; the editable canvas is the working record.
Measure effort per approved asset, including rejected attempts and corrections. Do not infer better ad performance from a visually stronger concept; evaluate the published campaign separately in the relevant advertising tools.
Start with one product and one image role
Download the worksheet, add approved references, and choose a single image role. Open Quickmake to explore the canvas, then sign in for account-based AI generation. Check current plans and credits before starting.
If the image is part of a wider campaign, use the ad creative brief to connect the visual with the audience, message, and review process. Explore AI image and video generation, AI image generation, and image editing.