@mentai yakuna 牛虎たつみ 2025 Style

Initial.The LoRA was trained on 99 images, consisting of:<ul><li>10 images: <code>year 2026</code></li><li>21 images: <code>year 2025</code>, including 2 cover illustrations</li><li>68 images: <code>year 2025, comic, monochrome</code>, including 4 images cropped from individual panels to focus on a single character</li></ul>The color illustrations are captioned with <code>artist name</code>, so this tag must be counteracted in the negative prompt.<a target="_blank" rel="ugc" href="https://tensor.art/models/1056710132529932182">Character reproduction</a> from the comics is decent, though not particularly accurate.dim32/alpha32, CAME, UNet LR 3e-05, cosine 990 steps, 99 images (1024²)

Files with SHA256: 4c1434943b4efcb10a125054b807f55f33642bdc27448a5d3c2c20a062ddb0e8