Mar 21, 2024

How to use Leonardo AI for consistent character images with image-to-image upload

In Short

For consistent character images with Leonardo AI, select the Absolute Reality model, set Image Weight to 0.7-0.9, Prompt Magic Strength to 0.4-0.5, and use detailed prompts including character traits and scenarios, employing the image-to-image upload feature for reference.

Key steps for consistent character creation

Model selection and settings adjustment:
- Choose the Absolute Reality model from Leonardo AI's "Finetuned Models" for high-quality, consistent outputs.
- Adjust Image Weight between 0.7-0.9 and Prompt Magic Strength between 0.4-0.5 for maintaining character consistency across images.
- Set init strength to 0.7 for similar results and 0.2 for more distinct outcomes when using Image2Image feature.
Prompt crafting and image guidance:
- Incorporate detailed references including comprehensive visual/textual descriptions of desired characters and scenarios.
- Utilize name anchoring by using specific names (e.g., Madonna, Pierce Sylvester) in prompts to ensure consistent character generation.
- For character uniqueness, develop prompts with distinctive traits and preferences, aiding AI in consistent portrayal.
Image-to-image upload specifics:
- Use the image-to-image upload feature by setting reference image weight appropriately for consistent new scene generation.
- Include specific attributes (e.g., clothing, personality) in prompts for recognizable, consistent AI-generated characters.
Iterative refinement and community insights:
- Engage with Leonardo AI user communities for additional tips and tricks on navigating character design challenges.
- Introduce subtle prompt changes to refine generated images, avoiding significant alterations for continuity.
Utilizing Leonardo AI tools for enhancement:
- Employ select, move, masking, and erasing features for detailed alterations, ensuring seamless AI model adaptation.
- Leverage undo/redo functionality for navigating design iterations, maintaining creative workflow flexibility.

Advanced techniques for character consistency

For dynamic character representations, integrate Leonardo AI with other tools like Midjourney AI for varied character outputs.
Train your own model with specific themes, mirroring images (768x768px) for Leonardo AI model training, enhancing consistency.
Create a custom model by naming a dataset after the character, uploading 10-15 reference images, and training on Leonardo AI.

Videos