<h1 id="deathbird-anima-character-lora">Deathbird - Anima Character LoRA</h1>Character LoRA for Deathbird, trained on Anima Base v1.0.<ul><li>Trigger: <code>deathbird</code></li><li>Type: Character</li><li>Base model: Anima Base v1.0</li><li>Recommended strength: 0.8-1.0</li><li>Training images: 65</li></ul><h2 id="identity-tags-(baked-into-the-trigger)">Identity tags (baked into the trigger)</h2>These describe the character and are present in every caption; you normally don't need to repeat them, but they help if a feature drifts:<code>1girl, white eyes, no pupils, purple lips, medium breasts</code>Negative (add to your negative prompt, for any outfit):<code>pale skin</code><h2 id="outfits">Outfits</h2>Add the tag set of the look you want (and its negative, when it has one, to the negative prompt):<ul><li>classic: <code>deathbird classic costume, pink helmet, pink horns, blue hair, feather hair, purple bodysuit, strapless, bare shoulders, silver harness, purple bracers, purple boots, thigh boots, high heel boots, pink cape, feathers</code> (41 imgs)<ul><li>negative: <code>glowing, aura</code></li></ul></li><li>spacesuit: <code>deathbird space suit, pink helmet, pink horns, purple bodysuit, turtleneck, long sleeves, purple gloves, silver harness</code> (24 imgs)<ul><li>negative: <code>yellow bodysuit, yellow clothes, aura, green theme, glowing</code></li></ul></li></ul><h2 id="mix-and-match">Mix and match</h2>The outfits above are how the training images were grouped. When generating, the parts below combine freely: start from the base, then pick one option of each group (and add its negative, if any):<ul><li>base: <code>deathbird classic costume, pink helmet, horns, purple bodysuit, strapless, bare shoulders, cleavage, silver harness, purple bracers, purple boots, thigh boots, high heel boots</code></li><li>cape:<ul><li>cape: <code>pink cape, feathers</code></li><li>no cape: (no tags) | negative: <code>pink cape, feathers</code></li></ul></li></ul><h2 id="usage">Usage</h2>Prompt with <code>deathbird</code>, then the outfit tags, then pose / scene / background. Captions were written generically (no names), so the trigger carries the identity.<h2 id="tips">Tips</h2><ul><li>Strength 0.8-1.0. Lower it (~0.5-0.6) when stacking with a style LoRA.</li><li>Combine with a style LoRA to place the character in another art style.</li></ul>

Files with SHA256: ef6ad582c11c40c5f7e883750fe81cdcd7c1770ea2e9069055dea1315cbcfcca