Start with a clear photograph of your product, upload it to Dreamina, describe the surrounding scene, and generate an image in the intended aspect ratio. Compare the product with the original before downloading the final file. For an existing item, a reference-led workflow gives you more useful information than asking AI to invent the product from text.
This tutorial follows an existing coffee-pouch example from source image to downloaded result. You will also learn when to use a background-only workflow, how to repair an inaccurate result, and how to turn one approved photo into a practical image set.
- Decide what kind of product photo you need
- Step 1: Take or choose a clear source photograph
- Step 2: Write a scene brief that keeps the product central
- Step 3: Upload the product and choose the composition
- Step 4: Generate the scene with a specific prompt
- Step 5: Review the downloaded result against the brief
- Step 6: Fix the problem with the right kind of edit
- Step 7: Export an approved master and prepare the image set
- Costs and permissions to check before a batch
- Frequently asked questions
Decide what kind of product photo you need
Choose the image’s job before choosing a style. An attractive lifestyle scene may work in a campaign while failing the requirements for a marketplace’s main listing image.
Dreamina is a practical starting point for reference-led lifestyle and campaign scenes. If the product must retain its original pixels, generate the surroundings separately and combine them with the original cutout in a layered editor. A product-specific background workflow is another option, provided you choose the mode that actually retains the foreground.
Step 1: Take or choose a clear source photograph
You can start with a phone photo. Use soft, even light, keep the whole item visible, and focus on the front label and distinguishing details. A plain background makes separation easier. Avoid glare that hides wording or deep shadows across important edges.
Photograph the exact color, size and packaging variant you intend to show. If you need a different view, provide a real photograph of that view. A generator may create a plausible unseen side, but plausibility does not verify its construction.
Keep the untouched source. For a real product, record a few identifying details: label wording, material, closure, number of parts and proportions. These become the acceptance checklist later.
The source below is a fictional metallic copper coffee pouch created for the existing demonstration. Its green label reads “FIELD NOTES COFFEE” and “250 g.” The valve, upper seal and side gusset give us additional checkpoints. It is not a customer SKU or a photograph proving a physical product’s specifications.
Step 2: Write a scene brief that keeps the product central
Describe what may change around the product separately from what must remain recognizable. Include the supporting surface, background, light direction, camera view and any props. Keep the first scene simple enough to judge.
For the coffee pouch, the brief was a pale kitchen counter in cool early-morning window light. One plain ceramic cup would sit farther behind on the right. The pouch needed to remain upright, front-facing and fully visible, with some space above and to the left.
That is more actionable than “make it premium.” It also gives you a clear reason to reject a result with several cups, a tilted pouch or an unreadable label, even if the image is attractive.
Choose a relevant setting without implying features the product does not have. A splash scene may imply water resistance; a fruit arrangement may suggest ingredients. Props should communicate the intended context accurately.
Step 3: Upload the product and choose the composition
Open the Dreamina AI image generator. Select an available image model and the workflow for using an image reference. Clear references from earlier work, upload the product source and verify the thumbnail before generating.
Choose a ratio that suits the image’s first destination. A 4:5 portrait composition can work for a product-centered social or campaign image; a horizontal banner needs a different balance of product and surrounding space. Choose the output setting available in your account and check the generation cost displayed before proceeding.
The existing coffee-pouch run used Seedream 5.0 Pro, a 4:5 ratio, the 2K interface setting and one reference. It produced one downloaded 1792 × 2240 JPEG under the account conditions recorded on October 27–28, 2026. This describes that sample, rather than promising that every account has the same model or output entitlement.
If exact pixel dimensions are required, treat those as a separate export task. A ratio such as 4:5 states the shape of the image, not its final width and height.
Step 4: Generate the scene with a specific prompt
Use the product source to identify the item, and the prompt to explain the new photograph. This was the executed prompt for the example:
Place the exact metallic copper coffee pouch from the reference on a pale stone kitchen counter in cool early-morning window light. A plain ceramic cup sits farther behind on the right. Keep the pouch front-facing, fully visible and upright, preserve every seam, fold, valve and the dark green label reading FIELD NOTES COFFEE and 250 g. Leave clean negative space above and to the left. Restrained premium product photography, no extra text, no people.
The preservation language sets a requirement; it does not lock the original pixels. The next step determines whether the output meets the brief.
For your own item, replace the product identity, setting and checkpoints. A serum bottle might require an unchanged cap, amber glass and an unobstructed label. A handbag needs accurate hardware and strap attachments. A chair needs the correct number of legs and a credible floor contact.
Generate a small set of candidates appropriate to your available credits. Save the original source and compare candidates against it. Avoid using the latest generated version as the only reference for the next attempt, because small product changes can accumulate.
Step 5: Review the downloaded result against the brief
The Dreamina result below placed the full pouch upright on a cool stone surface. The green label retained the required wording, and the valve, sealed top and side structure remained recognizable. The lighting and copper reflections suited the morning brief.
The fine folds were relit and regenerated, rather than preserved pixel for pixel. The background cup was partly cropped. The sample was useful as a recognizable product scene with those limitations; it does not prove exact packaging reconstruction or general performance across products.
Review your own image in two passes:
Approve product accuracy first. Strong lighting does not compensate for a rewritten label. Open the native downloaded image at a useful zoom; a small preview may conceal defects.
If the product has changed, follow a product-detail troubleshooting workflow before building more variations from it.
Step 6: Fix the problem with the right kind of edit
If the product is acceptable but a prop is wrong, target the prop or simplify the background. Request one cup instead of several, move it farther away, or remove it. Reinspect the product after the edit.
If the label or geometry changes, return to the original source. Make the preservation requirement more specific, but do not rely on repeated wording alone. If exact branding is essential, keep the authentic product or artwork as a separate layer and place it over the approved background in a controlled editor.
If the product appears to float, ask for a contact shadow consistent with its base and the light direction. If that repeatedly redraws the item, add the shadow on a separate layer instead. For reflective or transparent objects, an existing reflection can conflict with a new setting; a better source photograph or a physical shoot may be necessary.
If the composition lacks space for copy, extend or redesign the surrounding canvas rather than stretching the product. Keep headlines, offers and legal text editable in a design editor after the image is approved.
When a background-only route fits better
Photoroom’s AI Backgrounds workflow starts from a source image, removes its background and generates new surroundings. In the web interface, add the source, choose AI Backgrounds, then use a theme, image guidance or a custom background description. The Manual option accepts a free-form prompt. Review the generated scene, then download it or open it in the editor.
Choose the specific mode deliberately: Photoroom distinguishes AI Backgrounds from AI Scenes, which can relight and make small foreground changes, and Product Staging, which can change the foreground more substantially. AI Backgrounds requires AI credits and appropriate access; one feature’s availability does not establish another’s.
An existing run with the same coffee source retained the pouch and label in Photoroom AI Backgrounds, while its generated cups and lighting only partly matched the scene brief. It used different account conditions and a default two-image batch, so it is not a controlled platform-wide quality or cost ranking.
When Firefly’s image editor fits your workflow
Adobe’s documented product-shot route opens the Firefly image editor, uploads the product source, adds a scene prompt, chooses a ratio and available settings, then downloads a PNG. Style controls vary by model; the product-shot guide identifies particular style settings for Firefly Image 3 and 4. The app also hosts partner models, whose controls and usage conditions should be checked separately.
This is a documented alternative, not another product test performed here. Its ability to generate a new angle should not be treated as verification of an unseen product surface.
Step 7: Export an approved master and prepare the image set
Download the native approved image rather than capturing the preview on screen. Reopen the file, confirm its actual dimensions and format, and check the label, edges and crop again.
Keep a high-quality master. Create destination-specific copies from that approved file, leaving the product’s proportions unchanged. If you need an exact square or banner size, use a canvas with those dimensions and deliberately position or crop the master. Check whether the destination adds its own crop or compression.
Build a small set of images with distinct jobs:
- A clear main product view that meets the destination’s current image rules.
- Verified detail views from appropriate product photographs.
- A checked lifestyle image that adds context.
- A campaign layout with editable copy and a suitable crop.
For several scenes, keep the same source product and acceptance checklist. Change one scene brief at a time. A set is useful when each image answers a different buyer question, rather than repeating the same view with different decorations.
Costs and permissions to check before a batch
Check the selected model, generation cost, output settings and download conditions in the workspace. A daily credit allowance is not a fixed number of finished product photos: generations, revisions and different models can consume different amounts.
For commercial work, use inputs you are entitled to use and an actual model, feature and plan that permits the intended commercial use. Confirm the destination’s image requirements separately. Permission to generate an image does not establish that its product details are accurate or that every marketplace will accept it.
Frequently asked questions
Can I create a product photo from text alone?
Yes, for an imaginary product or concept. To represent an existing item, start from its actual image and verify the output. A text description cannot establish details that were never supplied.
Do I need professional camera equipment?
A clear phone photo can be a useful starting point. Sharp details, suitable light and a complete view matter more than an elaborate setup. Reflective, transparent or intricate products may need more careful photography.
Can AI replace every product shoot?
It can help create surroundings and campaign variations, but it cannot verify hidden geometry, actual fit or material behavior from an incomplete source. Use real photography where those facts determine the image’s value.
Can I turn the approved photo into a video?
You can use an approved image as a video reference in a supported workflow. Review the resulting motion separately: a good still does not prove that labels, shape and framing remain accurate throughout a clip.
Start with Dreamina for a clear reference-led product scene. Supply a source worth preserving, specify a simple environment, and approve the product before exporting the image set. When exact details require more control, retain the original product and change the surrounding scene through a suitable editing workflow.