Train Custom Models

Key benefits

• Character consistency

  Train character models for consistent appearance across any scene

• Unique art styles

  Create style models that capture your unique artistic aesthetic

• High-resolution training

  High-resolution training (1024px+) for sharp, detailed outputs

• Small dataset support

  Start with just 5–15 training images and scale up to 100 as needed

• Multiple base models

  Choose between Flux, Kontext, Bria or SDXL base models for optimal results

• Team sharing

  Once trained, instantly share models across your team or workspaces.

How it works

1. 1

   Build Your Dataset

   Upload 10–15+ high-resolution images (1024px+). For character models: ensure consistent features like hair, eye color, and skin tone across varied poses and expressions. For style models: maintain consistent visual aesthetics across diverse subjects and compositions.

2. 2

   Caption & Configure

   Add detailed captions describing key features, styles, and elements. Choose between Flux (faster, great for style consistency) or SDXL (more detailed, requires thorough captioning). Crop images to square format.

3. 3

   Train Your Model

   Click 'Train' and wait for completion. Training time varies by dataset size and base model. Your model will learn to reproduce the consistent character features or artistic style from your references.

4. 4

   Test & Optimize

   Start with simple prompts describing core features, then experiment with different scenarios. Use Prompt Spark for suggestions based on your captions. Pin your best results to improve future generations and refine the model.

5. 5

   Deploy & Organize

   Add a clear model description and appealing thumbnail. Tag your model or add to collections for easy access. Share with team members and start generating consistent, on-brand content at scale.

Use cases

• Character Model Training

  Create consistent characters across multiple scenes, poses, and expressions while maintaining core features like appearance and outfit

• Style Model Training

  Capture and reproduce specific artistic styles, color palettes, lighting techniques, and visual aesthetics across different subjects

• Brand Asset Generation

  Train models on brand-specific visual elements to ensure consistent marketing materials and promotional content

• Game Asset Pipeline

  Streamline creation of consistent game assets, characters, and environments that match your art direction

• Avatar & Portrait Creation

  Train models on personal photos to generate consistent avatar variations across different poses, expressions, and settings

• Concept Art Iteration

  Create repeatable concept art styles for rapid iteration and exploration of environments, creatures, and visual designs