In Dreamina, start with a clear photo of your product and a separate model reference, then describe how the person should wear or hold the item. Generate a small, specific scene and compare the result with both references before exporting. For a dedicated clothing workflow, Photoroom and Botika offer different ways to select a model and create on-model imagery.
First decide what “with a model” means. Holding a bag is different from dressing a model in a jacket, and both are different from measuring fit. AI can produce a plausible product image without establishing the garment’s real size, drape on a particular body, or a customer’s actual experience.
A model image can help you explore how a product fits into a fashion story. The fictional portrait and tote example below shows a reference-led creative workflow in Dreamina, from source selection to a new scene. Prices were checked on October 22, 2026.
- Choose the kind of on-model photo first
- 1. Prepare the product and model references
- 2. Upload the references and give each a role
- 3. Describe the interaction, not only the style
- 4. Generate, then compare the person and product separately
- 5. Correct one problem at a time
- 6. Finish for the intended destination
- When a fashion-specific workflow is a better fit
- Common questions
Choose the kind of on-model photo first
For broader tool selection before you commit to a workflow, see AI product-photo tools compared.
1. Prepare the product and model references
Use the actual product you need to represent. Start with a well-lit image that clearly shows its outline, color, materials, and distinctive details. For clothing, prepare the relevant front, back, and detail views where the selected tool supports them. A hidden fastening or an unseen back panel may otherwise be invented.
For the model reference, use an image you are entitled to use for this purpose. A fictional adult can be appropriate for a concept; an identifiable real person requires the relevant permission. Keep the person’s role honest in the finished image. A generated model is not a genuine buyer describing an experience.
In the existing Dreamina example, the two inputs were:
- A fictional adult woman with short curly dark hair and a blue shirt.
- A cream canvas tote with indigo handles and front pocket, a brass-colored snap, and visible stitching.
Write down those anchors before generating. It is easier to detect a changed pocket or missing fastening when you have already identified it as important.
2. Upload the references and give each a role
Use Dreamina’s image-generation workflow with the supported reference inputs. In this recorded case, the operator verified two references in a fixed order: the person first and the bag second. The chosen model was Seedream 5.0 Pro, with a portrait 4:5 composition and 2K setting.
Explain which reference supplies which information. The first should guide the person; the second should guide the product. This is more useful than uploading several pictures without saying whether they represent the item, the desired pose, or the background style.
Keep the first composition simple. A neutral studio setting and a straightforward holding pose expose product and contact problems more clearly than a crowded lifestyle scene. You can develop the background after the person and item work together.
Choose Seedream 5.0 Pro for the reference-led workflow shown here. A clear person reference and a separate product image let you direct the look and the accessory together.
3. Describe the interaction, not only the style
“Make a luxury fashion photo” leaves the most important questions unanswered. Say where the item sits, how it is supported, which face of it is visible, and which details need to remain.
The actual holding-a-tote brief was:
Show the same fictional adult woman in her blue shirt naturally holding this exact canvas tote by its two indigo handles at waist height, with the front pocket facing the camera. Three-quarter studio portrait, neutral warm-gray backdrop, soft fashion catalog light, relaxed believable posture. Preserve her face and hair, the bag shape, cream canvas, indigo pocket, brass snap and stitching. Correct hand-handle contact, no extra fingers, no text.
The result still needs comparison with the inputs.
For your own product, replace the interaction and detail list rather than reusing the bag instructions. A bottle held near the face, a shoulder bag, and a worn shirt each need different contact, visibility, and framing. Avoid asking the model to cover the feature you most need to show.
4. Generate, then compare the person and product separately
Here is the actual native 1792 × 2240 output from the recorded two-reference attempt:
The major visual anchors are recognizable: the blue shirt and short curly hair, cream canvas, indigo front pocket and handles, and brass-colored snap. The hands meet the handles in a plausible holding pose in this image. The operator accepted the result for the fictional catalog-style use.
However, the pose, folds, bag angle, and handle presentation were rendered into a new composition. That is not a pixel-identical transfer or a measurement of how a real bag would hang.
Review two things independently. First, does the model still match the intended reference closely enough for the project? Second, is the product accurate enough to represent the item? A convincing face should not distract from a changed logo, invented seam, or altered handle attachment.
Then inspect the interaction. Look for fingers passing through a handle, unsupported weight, a product fused into clothing, inconsistent shadows, or a pose that hides the item. Review the image at native size and at the size people will actually see it.
Keep the first scene focused on one pose and one product. Once the direction works, use the accepted image as a starting reference for useful variations.
5. Correct one problem at a time
If the person and product are broadly right but a small detail is wrong, make the correction specific. Identify the region and the desired change, and state which approved parts should remain. Compare the revised image with the source again; a local instruction can still affect nearby details.
If the product is substantially wrong, reconsider the source or composition before requesting more style changes. A clearer front-facing product photo, fewer overlapping fingers, or a less ambitious pose can make the task easier to judge. Repeatedly asking for “more realistic” does not tell the system which feature is inaccurate.
For clothing, pay particular attention to print placement, logos, seam positions, closures, neckline, sleeve length, and material appearance. Do not fix a visual problem by silently turning the garment into a different product. The product-detail correction guide explains how to separate an attractive rendering from a faithful item.
If you need an exact retained product layer, a compositing workflow may be more appropriate than regenerating the product with the person. The tradeoff is that realistic body contact and occlusion may require more manual work.
6. Finish for the intended destination
Keep an approved version before creating additional poses or backgrounds. For a product page, decide which image is the factual overview and which is the lifestyle or on-model view. A generated scene should not invent a feature that the source photography does not support.
Export the format and dimensions required by the destination, then review the actual file. A high-resolution setting does not remove an incorrect fastening or restore a detail the generation invented. For a set of images, compare the model, product, lighting, and color across the set rather than approving each picture in isolation.
The product-photo set workflow helps assign distinct roles to the overview, detail, and use-context images. A separate white-background guide is useful if the listing also needs a clean product-only view.
When a fashion-specific workflow is a better fit
Dreamina: flexible references for a creative scene
Dreamina is useful when you want to specify the person, product, interaction, and setting through references and a prompt. The tote example demonstrates that route for one fictional held accessory.
Advantages: Flexible scene direction, product and model references, and the ability to develop a selected visual through revisions.
Where it fits: Dreamina suits original fashion concepts and product-in-scene exploration. Direct the pose, styling, lighting and background together, then refine the contact between the person and accessory before building the next scene.
Free access and paid upgrade: Free accounts in the US receive 120 credits per day; daily allowances vary in other regions. Start exploring image ideas with Seedream 5.0 Lite and Nano Banana. For regular production, Dreamina Basic is USD 15 per month on monthly billing, with 1,575 monthly credits. Paid plans add watermark-free downloads and a faster generation queue. Credit use varies by model and settings.
Photoroom: dedicated AI Fashion Models controls
Photoroom’s official AI Fashion Models guide documents a more structured route. Supply at least one source image, select a premade model or upload a custom model reference, then choose the background, pose, quality, aspect ratio, and other available guidance. Multiple clothing or accessory images can support an outfit. After generation, review or edit the image and create variations where appropriate.
Advantages: Dedicated model, pose, and background choices can make recurring fashion-image production easier to organize. Standalone and supported batch or Shopify routes serve different workloads.
Limitations: AI credits are required, and settings affect consumption. Review the output before using it as a listing image.
Price and access: The price matrix lists Pro at USD 7.50 per month equivalent billed yearly, with 8,000 monthly AI credits and 1,000 exports; Max is USD 20.99 per month equivalent billed yearly, with 25,000 credits and 3,000 exports. Confirm the Fashion Models entry and required operation against the current tier and credit quote. Free is personal/noncommercial, so a Free export is not sufficient eligibility for a store’s commercial listing. Credits and exports are different allowances.
Botika: a fashion-focused service and correction process
Botika is relevant when the workload is a fashion catalog rather than one creative held-product image. Its supported workflow uses garment inputs and selections such as model, pose, and background. Do not assume it exposes the same open-ended prompt workflow as a general image generator.
Advantages: A fashion-focused process can fit a team that needs repeated on-model images and a defined correction route.
Limitations: Check the service and plan details. The availability of annual-plan photo fixes should not become a claim that every plan includes the same human retouching entitlement or that every generated garment is exact.
Price and access: The official FAQ describes eight trial credits without a card, with one credit per photo and five per video, and photo fixes on annual plans. The readable current pricing material contained conflicting Advanced-plan amounts, so this guide does not invent a definitive cash price. Confirm the selected plan’s billing and correction terms before budgeting a catalog. Trial credits describe an entry allowance, not a reliability result.
Common questions
Can I show a model wearing my clothing instead of holding a product?
Yes, use a workflow that supports that input and output. Supply clear garment views and inspect the generated construction, print, and fit appearance.
Can the image show customers what size to buy?
Use your actual size chart, product measurements, and appropriate real fit evidence. Do not infer sizing accuracy from an attractive on-model image.
Can I keep the same model across a whole catalog?
A reusable model reference or dedicated model selection can help guide consistency, but review the entire set. Face, hairstyle, body proportions, clothing, and product details can still drift. Test the difficult items before treating the workflow as a catalog standard.
What should I do if the model looks right but the product is wrong?
Return to the product reference and identify the exact error. Use a targeted correction if appropriate, or choose a workflow that retains more of the original product image. The model’s realism does not compensate for a misrepresented item.