# Objects ## Introduction Designing campaign visuals that accurately reflect your product’s aesthetic typically demands significant time, resources, and iterative revisions. Fermat streamlines this process by enabling you to **train the AI to understand your product's aesthetic**. This allows you to generate e-commerce images that align seamlessly with your brand’s style. This step-by-step guide will outline the process for leveraging Fermat’s advanced tools to produce professional-grade visuals that maintain consistency with your design objectives. This article will walk you through the process of using visual references of your product to craft designs that perfectly align with your aesthetic. ## Step by step guide To train an AI using Custom Styles, follow these steps: ### Create a Custom Style {% stepper %} {% step %} ### Accessing the Custom Style panel 1. Open Fermat and navigate to the *Custom Styles* section. 2. Click on *"New Style"* to begin the training process.

{% endstep %} {% step %} ### Upload visual references Upload at least 5 images showcasing your product from different angles. {% hint style="warning" %} Make sure your visual references are in high quality. Avoid pixelated or low-resolution images that could affect the accuracy of the AI-generated designs. If your dataset contains less than 5 images, you can upload the same images again to meet the required amount. {% endhint %}

To delete a specific visual reference, place the cursor over the image and press the *"Delete"* button. {% endstep %} {% step %} ### Set your Custom Style * Enter a name that represents the style you're training. * Select the **"Objects"** option. * Write a short description that explains your dataset. *e.g.* These are images of my "TerrainX sneakers".

{% endstep %} {% step %} ### Train the AI Click the "Start training" button. The AI will process your visual references and create a Custom Style within approximately **10 minutes**. {% hint style="info" %} The training time for your new Custom Style may depend on the number and size of the visual references you have uploaded. {% endhint %} {% embed url="" %} Watch this video to learn how to train the AI to generate images with your product's aesthetic {% endembed %} {% endstep %} {% endstepper %} ### Use Custom Styles {% stepper %} {% step %} ### Open the Custom Style

Press the Custom Style once the AI completes its training.

{% endstep %} {% step %} ### Enter your prompt {% hint style="info" %} Example: *A close-up shot of "TerrainX sneakers" on a person's feet, paired with beige athletic activewear leggings. The shoes feature a mix of mesh and textured material with a modern, sport-inspired design. The background is soft-focused, showing a blurred, spacious indoor setting with large windows, providing natural light. The person is in a relaxed pose, walking or standing, with a slightly blurred figure in the background. The overall vibe is stylish, athletic, and minimalist.* {% endhint %} {% endstep %} {% step %} ### Customize the settings Adjust parameters such as: * **Aspect ratio**: 1:1, 16:9, 9:16, 5:4, 4:5 * **Influence of AI on the original dataset.** This means that if the influence is 100%, AI will fully consider the dataset it was trained on. {% endstep %} {% step %} ### (Optional) Upload a reference image You can upload a reference image to restyle it using the unique DNA of your Custom Style. (Optional) To apply the essence of your Custom Style to a specific area of the reference image, simply **paint the desired area**.

Paint a specific area of your reference image to apply the essence of your Custom Style to that section.

{% endstep %} {% step %} ### Generate render Click the '*Generate*' button and let the AI process your inputs. In just a few seconds, your new design will be generated with the specified material applied.

{% hint style="warning" %} If the generated image doesn't perfectly match the dataset of your Custom Style, we recommend uploading the recently generated image in the 'Upload reference image' area. Then, paint the specific area to guide the AI in refining the design. {% endhint %}

This example shows how this tip makes a difference. The final result closely matches the provided dataset.

{% embed url="" %} Watch this video to learn how to generate e-commerce images that align seamlessly with your brand’s style. {% endembed %} {% endstep %} {% endstepper %} ### View dataset You can view your Custom Style's dataset by opening its own menu.

### Download dataset Can’t remember what your Custom Style images were trained on? You can now download them all with a single ‘Download’ button.

### Delete Custom Style You can delete your Custom Style by selecting the *"Delete"* option in its own menu.

## Examples Here are some examples of how designers use Custom Styles in Fermat:

*A sporty, "TerrainX sneakers" with a predominantly white design with beige details. The sneaker should have a sleek, modern look with textured, breathable mesh panels and a chunky, supportive sole. Pair the sneaker with knee-high black socks that feature a bold, colorful pattern—red, orange, yellow, blue—arranged in a gradient of vertical stripes, creating a striking contrast against the black background of the socks. The angle should focus on the side of the sneaker, highlighting the sole and the socks’ design, with the background remaining neutral and minimal to emphasize the product. The lighting should be clean and crisp, emphasizing the materials and design details of both the sneakers and socks.*

*A pair of light-colored "TerrainX sneakers" in pale cream with a textured mesh upper and chunky sole. Pair them with knee-high black socks featuring bold red, yellow, and blue stripes at the top. The person should be seated and the camera should focus on the shoes and socks. The background should be minimal with soft lighting highlighting the details.*

*A fashion photoshoot with a model sitting on a white cube. The model is wearing a simple, dark brown athletic crop top and matching biker shorts. They are accessorized with bold red sunglasses, white socks with black logos, and "TerrainX sneakers". The background is minimalist with a clean white backdrop, ensuring the focus stays on the model and the outfit. The model has curly hair and a relaxed pose, sitting with their hand resting on their knee. The lighting should be bright and clean to maintain a modern, sporty aesthetic.* ## Explore other types of Custom Styles {% content-ref url="moodboards" %} [moodboards](https://docs.fermat.app/custom-styles/moodboards) {% endcontent-ref %} {% content-ref url="materials-and-fabrics" %} [materials-and-fabrics](https://docs.fermat.app/custom-styles/materials-and-fabrics) {% endcontent-ref %} {% content-ref url="broken-reference" %} [Broken link](https://docs.fermat.app/custom-styles/broken-reference) {% endcontent-ref %} {% content-ref url="faces-and-people" %} [faces-and-people](https://docs.fermat.app/custom-styles/faces-and-people) {% endcontent-ref %} --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.fermat.app/custom-styles/objects.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.