Articles on: Advanced guides

Bringing several products into one image: 5 methods

A range shot, a complete ritual, a gift set, a collection staged together: as soon as one image has to bring several products together, each one must stay itself, with its shape, its label and its proportions. Seelab offers several ways to get there, from the fastest to the most controlled. This guide reviews them with a single example: the skincare range of the fictional brand Axène.


The common thread of all these methods: your Seelab models. A product that has its own model, trained on the platform, is reproduced faithfully whichever method you choose. The more of your range is modeled, the more reliable your multi-product images.


The 5 methods at a glance


Method

Best for

Product fidelity

Layout control

Volume

1. Multi-references in Creation

Exploring an idea quickly

Good on simple products

Through the prompt

One image at a time

2. Seelab model + multi-references

Everyday range visuals

Very good

Through the prompt

One image at a time

3. Composer

The visual that has to be perfect

Maximum

Full, down to the pixel

One composition at a time

4. Replace with your model

Adding or swapping a product in an existing visual

That of the model

Through the painted area

One image at a time

5. Workflow

Series: several settings, several formats

Very good

Through the prompt

Dozens of images per run


These methods combine: generate a scene with method 2, adjust a product with method 4, then scale up with method 5.


Why your Seelab models make the difference


A market generation model sees your products for the first time with every image. With several products in the same scene, it has to keep track of everything at once: it simplifies a label, swaps two bottles, changes a proportion.


A Seelab model has learned your product from your photos and your brief: shape, logo, text, materials, colors. In a multi-product image, it carries the product it knows and leaves the rest of the scene to the prompt and the references.


  • One model per product: the most reliable foundation. For a range of four products, four models.
  • Other products as references: a product without a model can enter the scene as a reference image, with slightly lower fidelity.
  • The model works everywhere: Creation, Replace, Workflow, and it strengthens blending in the Composer.


To get started: Create and train a model.


1. Multi-references in Creation


The most direct method. In Creation, add up to 10 reference images: your products, and optionally a setting. The prompt says what to do with each image.


The prompt and its four references


Place the three products from images 1, 2 and 3 side by side on the travertine shelf from image 4, well spaced: the mist from image 1 on the left, the serum from image 2 in the center, the oil from image 3 on the right. Keep their shapes and labels exactly. Keep the setting, the lighting and the leaf shadows from image 4.


The result


A few tips:


  • Sharp photos on a plain background, one per product: details are better preserved.
  • Give each image a role and a place: "the bottle from image 2 in the center".
  • Say what must not change: shape, label, proportions between products.


Its limits: the more products there are, the higher the risk that a detail slips, especially small text. It is the right method to explore an idea; for a final visual, move on to method 2 or 3.


In detail: Using up to 10 reference images.


2. Your Seelab model + references


The same thing, with a Seelab model on top. Choose the model of one of your products in the model selector, under Custom models, then add the other products as reference images.


The model carries its product without you having to describe it; you can focus the prompt on the scene and on where each product goes.


For the example: the Axène Bakuchiol Serum model, with the three other products of the range as references. Three prompts, three settings.


The model, the three references and three results


A prompt along these lines:


The complete Axène range lined up on pale travertine plinths: the Bakuchiol serum, the Ceramide Cream jar from image 1, the Centella Mist spray from image 2 and the Squalane Oil pump bottle from image 3, each product identical to its reference. Morning light, leaf shadows on the wall, a few centella leaves and seeds scattered around.


A few tips:


  • Pick the most demanding product as the model: the one with the most text, the most complex shape, or the hero of the image.
  • Name each product in the prompt and link it to its image.
  • Vary the scene freely: bathroom, flat lay, nature. The model keeps the product stable from one image to the next.


Is the whole range modeled? Use the hero product's model, and for the others, the dataset photos of their models as references: they are available directly in the Resource Hub.


3. The Composer: pixel-perfect composition


When placement matters as much as the product, the Composer (Edition › Composer) lets you build the scene yourself, layer by layer, then harmonizes it.


The Composer: the Axène range on travertine plinths


  1. Add the setting, then one layer per product (1) from the Resource Hub: your imports, your collections or the visuals of your Seelab models. Place and resize each product on the canvas.
  2. Blend each product: Seelab creates a Fusion layer (2) that integrates the product into the scene, with the setting's light, reflections and shadows (up to 7 credits per fusion). Retry offers another version.
  3. Send the composition to Edit, Reframe, Magnify or Video with Use composition in (3).
  4. Export with Download all (4), as a .ZIP or a layered .PSD for your studio.


Product reconstruction works better with a Seelab model. During blending, Seelab reconstructs the product in the light of the scene. When that product has its own model trained on the platform, the reconstruction relies on what the model knows about the product: the logo, the label text, the materials and the proportions are better preserved. The difference shows most on small text and transparent bottles.


A few tips:


  • Start from good product images: cut out, sharp, at an angle consistent with the setting. Remove background cuts out a layer in one click.
  • Blend one product at a time, starting with the largest or the one in front.
  • Keep realistic scales between products: a cream jar is wider than a serum.


In detail: Composer: create a composition and Composer: layers, blending and PSD export.


4. Replace, with your model


Already have a successful visual and want to add a product, or swap one? In the editor, Replace combined with your Seelab model inserts the exact product where you paint.


  1. Open the image in the editor and choose Replace.
  2. Paint the spot, slightly wider than the product to insert.
  3. Choose your model with Select a model, to the right of the prompt bar.
  4. Describe the placement: where and how the product sits, and the lighting.


Before: the central serum is painted, the Ceramide Cream model is chosen

After: the cream jar has taken its place


The open Axène Ceramide Cream jar, sitting on the travertine shelf between the two bottles, same lighting and same leaf shadows


Useful for:


  • building a scene product by product, keeping what already works;
  • rolling out the same visual for every product in the range;
  • fixing a product that another method rendered poorly.


In detail: Insert your product with your Seelab model.


5. The workflow: the range at scale


When you need the same range in ten settings or five formats, the Workflow does the job in a single run. The principle: a Reference image node holding your products, one prompt per setting, and generation nodes with your model.


The workflow: references, prompts and generations


  1. A Reference image node (1) with your products, set to All together: each generation receives all the references at once.
  2. One Prompt node per setting (2): bathroom, beach, forest… The start of the prompt is the same, only the scene changes.
  3. Image nodes with your Seelab model in the selector (3), and the format you need.
  4. Run all (4) produces every image; a collection node can file them automatically.


For the example: the Axène Bakuchiol Serum model, the three other products as references, three settings, in 16:9 and 2K. 19 credits for the three images.


Three settings in a single run


A few tips:


  • Write a shared prompt base that names each product and links it to its image, then add the scene.
  • One image is off? Rerun that node alone, without rerunning the whole workflow.
  • For another range, duplicate the workflow and change the references and the model: nothing else changes.
  • One reference per generation? The One by one setting creates one generation per reference image: useful to roll a scene out product by product.


In detail: Component reference and Generate a workflow with AI.


Which method to choose?


  • You are looking for a staging idea: method 1, then move to method 2 once the direction is set.
  • You produce regular range visuals: method 2, with the hero product's model.
  • The visual must be flawless, for a campaign, print or a gift set: the Composer, with modeled products.
  • The visual already exists and a product is missing: Replace with your model.
  • You have dozens of visuals to produce: the workflow.


Cheat sheet


  • Model your products: that is what makes every method reliable.
  • One image per product, sharp and on a plain background, when you use references.
  • Give each image a role in the prompt and say what must not change.
  • The Composer for full control, one product blended at a time; reconstruction is better with a Seelab model.
  • Replace to add or swap a product in an existing visual.
  • The workflow for series: references "All together", one prompt per setting.


With the Seelab team


A whole range to stage? As part of its onboarding and support, the Seelab team can model all your products, check each model's fidelity, build your range-visual workflows and train your teams. Talk to your Seelab contact or write to us from the chat.


Learn more


Updated on: 12/11/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!