Usage info<ul><li>Base model: Animagine v3.1</li><li>Activation tag: shibuyakanon</li><li>Recommended strength: 0.9</li></ul>This model comes in two different variants at two different epochs:<ul><li>56 epochs: capture details very well (including Kanon's different appearances such as school uniform with/without jacket, training outfit, at home with glasses, ...).</li><li>46 epochs: capture general details well but perform better if generalization is prioritized over accuracy.</li></ul>Other tags for appearance/outfit:<ul><li>School uniform with jacket: kanonuniform, school uniform, blue jacket, pinafore dress, red ribbon</li><li>Uniform without jacket: kanonuniform, school uniform, white shirt, pinafore dress, red ribbon</li><li>At home: kanoncasual, no bangs, (hair bun:0.8), short hair, glasses</li><li>Training hairstyle: kanontraining, low-tied hair, hair band, blue hair band</li><li>Training outfit: kanontraining, two-tone shirt, white t-shirt, red sleeves, short sleeves, off-shoulder shirt</li></ul>Training details:<ul><li>Dataset: 160 cropped images, anime only</li><li>Total number of steps: around 3000 (56 epochs)</li><li>Config: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/1771/how-to-create-near-perfect-character-and-style-loras-for-sdxl-10-latest-update-07092023">How to create near-perfect character and style LoRa's for SDXL 1.0 | Civitai</a></li></ul>

Files with SHA256: d522417dd44b97b760a9a698c8cfba6bf2cb946cc95cf22fcc2866ba8d74b6f0