<h2 id="model-overview">Model Overview</h2><p>This LoRA was trained on a dataset of 200 hand-picked images, captioned using a mix of Gelbooru tags and curated natural language. It is highly flexible and works seamlessly alongside other character LoRAs.</p><h2 id="usage-and-settings">Usage and Settings</h2><p>For the best results, include terms like "indoors" or "outdoors", along with "blurry background" in your positive prompt.</p><p><span style="font-family:Google Sans Text, sans-serif;line-height:1.15"><strong>Trigger Word: </strong>@b33d34dfl0w</span></p><p><span style="font-family:Google Sans Text, sans-serif;line-height:1.15"><strong>Weight: </strong>0.8 – 1.0</span></p><p><span style="font-family:Google Sans Text, sans-serif;line-height:1.15"><strong>Portrait (2:3): </strong>1024 x 1536</span></p><p><span style="font-family:Google Sans Text, sans-serif;line-height:1.15"><strong>Landscape (3:2): </strong>1536 x 1024</span></p><h2 id="known-issues">Known Issues</h2><p><strong>Warning:</strong> The model tends to bleed Japanese text into the generated images. Adding "Japanese text, text" to your negative prompts clears this up, but at the cost of slightly weakening the core art style. Check the preview images featuring Marin Kitagawa to see the exact effect.</p><h1 id="if-you-are-the-original-artist-and-would-like-this-model-removed-please-send-me-a-direct-message."><strong><u>If you are the original artist and would like this model removed, please send me a direct message.</u></strong></h1><p></p>

Files with SHA256: dbe3c9e2d5a28fe00142da2f0bf19fb05a9b5d4fd5fa4841627ed313420330b4