> ## Documentation Index
> Fetch the complete documentation index at: https://docs.krea.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Training

> Train custom Krea models on your own datasets to preserve style consistency across campaigns, characters, products, and long-running creative work.

# Training Models

The **Training tool** in Krea allows you to **train AI models on custom datasets**, ensuring consistency across projects. This is useful for **brand identity, character design, and stylistic continuity**.

## Key Benefits

* Create consistent visual styles across multiple generations
* Develop custom character models that maintain recognizable features
* Establish brand-specific aesthetics for marketing materials
* Save time by training the AI to understand your unique requirements

## Steps to Train a Custom Style in Krea

### 1. Upload a Dataset

* Users must upload at least **3 images** of the same **art style, character, or object** for AI training.
* Larger datasets (10-30 images) improve the model's ability to generalize.
* For best results, include varied examples that showcase the key elements you want the AI to learn.

### 2. Generate a Style Code

* Once trained, Krea assigns a **unique style code** that can be applied to **Flux, Edit, and Enhancer outputs**.
* Example: Training on **hand-painted watercolors** will allow Krea to replicate that style on any input.

### 3. Apply & Refine the Style

* Apply the trained style to new generations.
* Refine the model by uploading additional images for more accuracy.
* Publish styles for broader application (optional).

## Best Practices for Training Jobs

* **Curate a consistent dataset** with uniform lighting, color balance, and resolution.
* Start with **simpler styles** (e.g., digital paintings, graphic designs) before training highly detailed textures.
* Keep refining the dataset over multiple iterations for improved results.
* Use images with clear, distinctive features that represent the style you're trying to capture.
* For character models, include various poses and expressions to help the AI learn the core attributes.
* Balance variety and consistency in your training data for the most versatile results.

## Applications for Trained Models

### Brand Identity

Train models on your brand's visual assets to maintain consistent aesthetics across all AI-generated content.

### Character Design

Create consistent characters for animations, games, or storytelling by training the AI on your character designs.

### Artistic Styles

Capture your unique artistic style or emulate specific techniques to apply across multiple projects.

### Product Visualization

Train models on your product catalog to generate consistent product images in various contexts.

## Using Trained Styles Across Krea

Once you've created a trained style, you can apply it in various ways:

* As a style reference in **Flux** for new image generations
* As a guidance for modifications in **Edit Mode**
* As a style influence during upscaling in **Enhancer**
* As a consistent aesthetic for video generation