Articles on: Advanced guides
This article is also available in:

Seelab product model: one model for your images, videos and workflows


A Seelab product model learns your product once, from a few photos and a brief. From then on, it works everywhere in the platform: to generate images and videos, to retouch and upscale, and to produce in series in a workflow. This guide follows a single product from start to finish: the High Tide salted butter caramel cream from the fictional brand Kervala.


From brief to model, from model to all your visuals


What you can do with a product model


Where

What you do

Observed credits

Creation

Images of your product in any scene

8 per image (1:1, 2K)

Video, Custom Models tab

Your product in motion, its label intact

24 for 4 seconds (1080p, with sound)

Edition, Replace tool

Your product inserted into an existing photo

shown before you launch

Edition, Magnify tool

An upscale that keeps your product's details

shown before you launch

Workflow, Image node

A series of visuals, launched in one click

8 per image (1:1)


1. What the model learns: photos and a brief


The High Tide model was trained with 4 photos of the jar and a 6-page PDF brief. The photos show the product. The brief says everything the photos don't:


  • the house style guide: logos, palette, typefaces, photo style;
  • the product sheet: name, ingredients, how it is made, dimensions, serving suggestions;
  • the reference views and pack anatomy: the lid and its emblem, the anchor guarantee band, the label, the scale.


The Kervala High Tide PDF brief


On the model page, in Custom models:


The model page


  1. Dataset: the training photos.
  2. Briefing documents: the PDF attached to the model.
  3. Use model: opens Creation with this model selected.


How to build this brief: The training brief: text, PDF or both.


2. The brief does the work: short prompts, accurate results


This is where a product model makes the difference. The prompt only describes the scene. The model knows the rest, because the brief taught it.


Four short prompts, four facts taken from the brief


  • "Surrounded by its ingredients": the model brings in butter, cream, sugar and Guérande salt. That is the ingredient list from the product sheet, without writing it in the prompt.
  • "In a hand": the jar is the right size, 8 cm across, as stated in the brief.
  • "Top view": the lid is exact, with its emblem, the sailing boat, LE CROISIC · BRITTANY · 1934 and the gold anchor band.
  • "Served the way it is meant to be enjoyed": the model picks crêpes, the first serving suggestion on the product sheet.


In all four images, the label, the logo and the colors stay those of the product. This level of context and compliance is what lets you produce fast without checking every visual line by line.


A few tips:


  • Describe the scene, not the product: the model already knows its shape, its label and its details.
  • Put the facts in the brief once and for all: ingredients, dimensions, uses, what to avoid. You no longer need to repeat them in your prompts.
  • A detail drifts? Complete the brief or add a photo, then Retrain from the model page.


3. In images


In Creation, pick your model in the model selector, under My custom models, then describe the scene.


Creation with the High Tide model


  1. The model: Kervala - High Tide caramel.
  2. The prompt: the scene only.
  3. Generate: the cost is shown on the button, here 16 credits for two images.


In detail: Using your model in image and video.


4. In video


The same model animates your product, with no new training. In Video, Custom Models tab:


The Video space, Custom Models tab


  1. Custom Models: the tab for videos with your Seelab models.
  2. Your model: here Kervala - High Tide caramel.
  3. Settings: format, resolution, duration, audio, number of videos.
  4. Video prompt: the scene and the motion, and what must not move.
  5. Generate: 24 credits for 4 seconds in 1080p, with sound.


Breakfast table by the sea, next to a plate of crêpes. A golden spoon slowly drizzles caramel over the crêpes. Slow, gentle camera push-in toward the jar, soft morning light, light breeze in the linen napkin. The jar and its label stay perfectly still and sharp.


The generated video, without sound


The shot, from the first to the last second


The caramel flows, the camera moves in gently, and the jar stays sharp from start to finish, label included. For a product video, that is what matters: the product never distorts.


In detail: Animate your product with your video model.


5. In the editor: insert and upscale


Insert your product with Replace


You have a great photo, but without the product? Replace puts it there, with your model. This is inpainting: only the painted area changes, the rest of the photo stays identical.


The Replace tool with the High Tide model


  1. Replace: the tool, in the left bar of Edition.
  2. The area: paint where the product should appear, at its size.
  3. The prompt: where the product sits and how it blends into the scene's light.
  4. The model: pick your Seelab model.
  5. Launch.


Before, the area, after


The jar takes its place next to the crêpes, with the morning light and a soft shadow, and the scene around it doesn't move.


In detail: Insert your product with your Seelab model.


Upscale with Magnify


Magnify enlarges the image while adding details. With your model selected, those details are your product's.


The Magnify tool with the High Tide model


  1. Magnify: the tool, in the left bar.
  2. Parameters: Upscale x1, x2 or x4, and the Preset. Fidelity stays closest to the original image.
  3. The prompt: the materials to render (textured paper, gold foil, glass, caramel).
  4. The model: your Seelab model.
  5. Launch.


The label, before and after Magnify


On the label, the small lettering is corrected: "CANNERY & BISCUIT HOUSE" becomes readable again, as on the real jar.


Choose Fidelity for a packshot or a label. Keep the other presets for images where you accept that materials get reinterpreted. Always check the result by zooming in before using it.


In detail: Upscale (Magnify) an image.


6. In a workflow: produce in series


The model also plugs into a workflow. Here, three visuals for the brief's three photo families, launched in one click.


The workflow with the High Tide model


  1. CSV Prompts: one row per visual to produce.
  2. Prompt rows: three short prompts, as in Creation.
  3. The Image node: with the Kervala - High Tide caramel model, in 1:1, one generation per prompt.
  4. Add to Collection: the visuals are filed automatically in the brand's collection.
  5. Output: to review and download the batch.
  6. Run all: 24 credits for the three images.


Three prompts, three visuals, the same product


To get one image per prompt, use a CSV Prompts node. Several Prompt nodes connected to the same Image node are joined into a single prompt.


To go faster, describe what you need in Generate your workflow: Seelab builds the canvas, and you just check the model and the prompts. See Generate a workflow with AI and Seelab automations.


Cheat sheet


  • One model, the whole platform: Creation, Video, Replace, Magnify, Workflow.
  • The brief drives compliance: ingredients, dimensions, pack details and uses are learned once.
  • Short prompts: describe the scene, the model takes care of the product.
  • In video, the product and its label stay sharp.
  • In the editor, insert your product into an existing photo and upscale it without altering it.
  • In a workflow, go from one visual to a series.


With the Seelab team


Modeling an entire range, writing the briefs from your product sheets, validating compliance with your teams: this is the core of our onboarding and support. For food brands, see also Seelab for food. Talk to your Seelab contact or write to us in the chat.


Learn more


Updated on: 17/11/2026

Was this article helpful?

Share your feedback

Cancel

Thank you!