Train Model

The Custom Trained Models feature empowers you to train AI with your data, giving you control over style, faces, and objects in your generated images. Create truly unique visuals tailored to your exact creative vision!

Key Benefits:

• Unprecedented Control: Shape the style, faces, and objects in your images.

• Unique & Personalised Content: Generate visuals impossible with standard AI.

• Creative Empowerment: Unlock new artistic and design avenues.

• Increased Efficiency: Automate visual content creation.

Training Variants:

• Object Training: Teach the AI to recognise and generate specific objects (products, items). Ideal for e-commerce, brand marketing, and product visualisation.

• Character Training: Create consistent characters with realistic facial features that maintain identity across generations and poses. Perfect for character development and realistic face editing.

• Style Training: Capture unique artistic aesthetics and apply them to any new prompt. Great for designers and artists seeking visual consistency.

How It Works:

1. Upload Images & Settings: Upload a collection of images representing your desired object, face, or style. The images should fulfil all the requirements.

2. Model Training: Our platform trains a custom AI model using your dataset.

3. Image Generation: Generate new images using your trained model within the Text to Image feature.

Tips for Getting the Best Results from Training

The quality of your dataset has the biggest impact on your model’s performance. A thoughtfully curated dataset ensures your model learns correctly and produces high-quality outputs.

General Guidelines

• 20–30 images is the ideal range per model.

• Images should be clear, high resolution, and consistent in framing and subject size.

• Avoid blurry, cluttered, or low-quality inputs—these will directly reduce output quality.

Note: Images must be at least 512 × 512 pixels in resolution. Uploading smaller images may result in poor training outcomes or failed uploads.

If you’re training a Style

• Include a range of content types (e.g. people, objects, landscapes), all reflecting the same artistic style.

• The model learns how the style is applied, not just what the subject is so diversity in content helps it generalise better.

If you’re training a Character or Object

• Use images from multiple angles: front, back, and side, views.

• Include the subject in different settings or backgrounds to avoid overfitting to a single scene.

• Keep the subject size consistent across images for more accurate generation. (Object only)

• Make sure the subject is clearly visible and not obstructed.

Why these matter

If the dataset lacks variation or consistency, your model may struggle to learn and generate reliably.

A good dataset helps the model understand what’s important. Whether it’s style, identity, or object shape without confusing signals.

Think of your dataset as the blueprint: the clearer and more complete it is, the better your AI will perform.

We're excited to see what you create with Custom Trained Models! Explore new creative possibilities with ImagineArt!