This is a high-quality LoRA trained for Marisa Kirisame, the beloved character from the Touhou Project. The illustrious in the model's name signifies its core purpose: to generate exquisite, magnificent, and highly detailed illustrations, going far beyond simple character replication.
Illustrious Quality
This model was trained on a curated dataset of high-quality illustrations, excelling at generating images with detailed shading and rich details. From the strands of her hair and the folds of her clothing to the glow of her magic, every element is rendered with remarkable fidelity.
Excellent Outfit Generalization
The training data includes a variety of Marisa's classic and alternative outfits. This means you are not limited to her iconic black-and-white witch attire. You can easily prompt for different costume designs, and the model will understand and render them beautifully, showcasing its strong generalization capabilities.
Native High-Resolution Support
Trained on 832x1216 resolution images with ARB (Aspect Ratio Bucketing) enabled, this model is inherently skilled at handling portrait-oriented compositions. This significantly reduces issues with proportions and composition at high resolutions, making it perfect for generating detailed half-body and full-body shots.
Natural Language Driven
The dataset was tagged using descriptive, full sentences, enhancing the model's ability to understand natural language prompts. You can write your prompts as if you were describing a painting, allowing for more precise control over the scene, pose, and atmosphere.
Recommended Weight:
0.7 - 1.0. A weight of 1.0 provides a strong character likeness, while a slightly lower weight can improve compatibility when blending with other LoRAs.
Core Prompts:
kirisamemarisa, blonde hair, yellow eyes, witch hat
(Feel free to change yellow eyes to brown eyes or blue eyes; the model responds well to these variations.)
Style & Quality Prompts:
For the best illustrative results, consider adding:
masterpiece, best quality, digital art, detailed shading, anime style
Recommended Resolution:
Start with 832 x 1216 or other similar high-resolution aspect ratios (e.g., 768 x 1024). Thanks to ARB, the model adapts well to various dimensions.
High-Resolution Training: Trained on an 832x1216 dataset for native support of high-res portrait images.
Excellent Outfit Generalization: Not overfitted to the classic costume; easily customizable via prompts.
Illustrative Style: Focused on generating images with rich detail and sophisticated shading.
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
