Skip to main content
All CollectionsSuperstudio Core Features
Getting Started with ModelMaker
Getting Started with ModelMaker

ModelMaker is a game-changing feature that lets you train custom AI models (LoRAs) that capture your unique style, characters, and creative vision.

Updated this week

The Basics

What is ModelMaker?

ModelMaker is your personal AI trainer. Upload a set of images showing what you want to create—whether that's a character, place, style, or object—and ModelMaker will create a custom model that allows you to generate new images that match your vision.

What can I create?

You can create your own custom models based on four different training types:

  1. Character: perfect for characters, portraits, and consistent faces. You can train on photos of yourself to create endless profile pictures in any style or setting imaginable.

  2. Place: bring locations and environments to life. For example, train on photos of a cozy coffee shop to visualize it in different seasons, times of day, or even fantastical settings.

  3. Style: capture your artistic style or visual aesthetic. Upload your artwork and let the AI learn your unique creative vibe - then apply it to anything you can imagine.

  4. Object: focus on specific items or products. Train on your product photos to create marketing materials, product variations, or placement mockups.

How to Use ModelMaker

Step 1: Prepare Your Training Data

  1. Determine your model type (Character, Place, Style or Object)

  2. Gather your training images. Follow these best practices:

    • Character/Style: 25-40 images recommended

    • Place: 5-25 images recommended

    • Object: At least 5 high-quality images

  3. Ensure the images meet technical requirements:

    • JPG or PNG format

    • Max 20MB per image

    • Clear, well-lit, and in focus

    • No blurry or poor quality shots

    • Consider variety within consistency (angles, lighting, contexts

  4. Add the images to your Canvas and group them into a Collection

  5. You can reorder the images and name your Collection to keep things organized

Step 2: Train Your Model

  1. Open the Explore menu and click “Co-Create”

  2. Add ModelMaker to your Canvas

  3. Create a new Collection—or drag in an existing one—and connect it to the ModelMaker Flow

  4. Choose your training type (Character, Place, Style or Object)

  5. Enter a unique keyword for your model

  6. Click “Generate” and wait for completion (typically a few minutes)

Step 3: Generate with Your Model

  1. Add Image Lab 2.0 to your Canvas

  2. In Image Lab 2.0:

    • Open the Custom Models menu

    • Select your trained model(s)

      • Tip: you can add up to 3 different models at once

    • Add a prompt to the Subject Element that includes your keyword (optional: add a subject reference image)

    • Adjust the settings as needed

      • Tip: if you have multiple models connected, you can use the weights slider to effect which model(s) influence your outputs the most

  3. Generate and explore!

  4. You can create new Collections for your various outputs to keep things organized

ModelMaker Training Data Tips

For all models:

  • Minimum 5 images required

  • Use clear, high-quality images

  • Remove blurry or poor-quality shots

For Character models:

  • Mix close-ups and full body shots

  • Include various expressions

  • Show different angles

  • Keep lighting diverse

For Style models:

  • Use finished artwork only

  • Include subject variety

  • Maintain consistent style

  • More images = better results

For Place models:

  • Capture different viewpoints

  • Include details and wide shots

  • Show varying conditions

For Object models:

  • Show multiple angles

  • Include size context

  • Use clean backgrounds

  • Highlight key features

Conclusion

ModelMaker is your tool for creating personalized AI models that match your creative vision. Start with these basics, then experiment and explore. Remember, better training data leads to better results!

Need more help? Contact our support team or join our Discord community for guidance.

Did this answer your question?