💬 Discord: <a target="_blank" rel="ugc" href="https://discord.gg/reMJwfNp3j">join here</a> · ☕ Ko-fi: <a target="_blank" rel="ugc" href="https://ko-fi.com/techannel">support the factory</a><h1 id="typhoeus-arknights:-endfield">Typhoeus 提丰 — Arknights: Endfield 明日方舟：终末地</h1>Horns, drills and a tail. All of it rides the trigger.<hr />Character LoRA for Typhoeus, trained with a baked identity. The trigger alone carries the canonical look: the lavender twin-drill hair, the pink eyes, the curved black horns and their headgear, and her white-and-black rig with the red emblem. None of it needs prompting.Her outfit pieces stay tagged, so they remain steerable. That cuts both ways — name them when you want them exact, drop them when you want her in something else.<h2 id="prompt">📋 Prompt</h2>Default look. The trigger on its own already gives the full canon rig:<pre><code>typhoeusendfield, 1girl, solo, cowboy shot, looking at viewer</code></pre>Pinned canon kit, when you want every piece exact:<pre><code>typhoeusendfield, 1girl, solo, horn hairband, white thighhighs, detached sleeves, black gloves, bare shoulders, tail</code></pre>With her bow:<pre><code>holding bow (weapon), weapon</code></pre>Negative, worth keeping:<pre><code>chibi, speech bubble, english text, multiple views, watermark, signature, meme</code></pre>NO quality tags needed. Keep the rest lean, long tag stacks dilute the trigger.<h2 id="version">🧩 Version</h2><ul><li>🔵 v1.0 Anima, for Anima / Qwen-Image DiT checkpoints. Weight 1.0 (usable range 0.85 to 1.0).</li><li>🟢 v1.0 Illustrious, for SDXL / Illustrious / Pony / NoobAI checkpoints (showcases rendered on WAI). Weight 1.0 (usable range 0.85 to 1.0). Same trigger, same prompt stacks.</li></ul><h2 id="tips">💡 Tips</h2><ul><li>Small, dense dataset. This is a 16-image bake — the identity is strong, but she has fewer poses and angles in her than a big LoRA. If a pose fights you, say it explicitly rather than rerolling.</li><li>Name the facing. Scene prompts without <code>looking at viewer</code> will sometimes turn her around — several training shots are back or three-quarter views. <code>facing viewer</code> settles it.</li><li>Never negative her colours. <code>purple hair</code>, <code>pink eyes</code> and <code>horns</code> were stripped from every training caption, which is why they ride the trigger. Negating them fights the LoRA.</li><li>The tail is canon kit, not a bonus — <code>tail</code> keeps it in frame on full-body shots.</li><li>Alt outfits are easy. The clothes are tags, not the trigger. Dress her in anything and the horns, hair and face hold.</li><li>Keep <code>solo</code> in. A few training images are group shots, so without it company turns up.</li></ul><hr />More of my LoRAs: <a target="_blank" rel="ugc" href="https://civarchive.com/user/Techannel">Techannel on Civitai</a>. Likes and Buzz keep the factory running ❤️

Files with SHA256: dd3902c81831caa1085d2194b475c5d5004f577fa6e16570199b88ceb218c67e