<a target="_blank" rel="ugc" href="https://www.patreon.com/cw/Techannel"><img src="https://cdn.jsdelivr.net/gh/Techannel/assets@v3/banners/patreon-banner.png" alt="Early access on Patreon: all early-access LoRAs free for 7 days, then €5/month / 早期アクセスのLoRAすべて、7日間無料でお試し" /></a><a target="_blank" rel="ugc" href="https://techannel.app/"><img src="https://cdn.jsdelivr.net/gh/Techannel/assets@v3/banners/archive-banner.png" alt="Techannel archive: see what is in early access and vote on the next LoRA, free / 早期アクセス中のLoRAを見る・次のLoRAに無料で投票" /></a>💬 Discord: <a target="_blank" rel="ugc" href="https://discord.gg/TYHy7fpEXx">join here / 参加はこちら</a> · 🔓 Patreon: <a rel="ugc" href="https://www.patreon.com/cw/Techannel">new LoRAs 7 days early / 新作LoRAを7日早く</a> · ☕ Ko-fi: <a target="_blank" rel="ugc" href="https://ko-fi.com/techannel">support the factory / 応援する</a> · 💼 <a rel="ugc" href="https://ko-fi.com/techannel/commissions">Request Commission / コミッション依頼</a><h1 id="rikuhachima-aru-3-outfits-blue-archive">Rikuhachima Aru 陸八魔アル, 3 Outfits — Blue Archive ブルーアーカイブ</h1>Her face comes free. The clothes are the part you ask for.<hr />One file, three canon outfits, a trigger for each. Her identity is baked into all three: the pink hair, the horns, the halo, the eye colour. None of it needs prompting.The clothes I left tagged on purpose, so they stay steerable. That cuts both ways. Each trigger needs its outfit words to land its look. I checked: <code>arunewyearba</code> on a bare prompt came back in the red gown. With the kimono words in, it is exact.Start from these.<h2 id="prompt">📋 Prompt</h2>Default look. White shirt, red ribbon, fur coat off the shoulders:<pre><code>arubaseba, white shirt, collared shirt, neck ribbon, red ribbon, black skirt, high-waist skirt, fur-trimmed coat, coat on shoulders, white gloves</code></pre>New Year kimono. Rose print, black fur collar, hair up:<pre><code>arunewyearba, kimono, red kimono, print kimono, floral print, wide sleeves, obi, fur-trimmed kimono, hair flower, hair bun</code></pre>Red gown:<pre><code>arudressba, red dress, strapless dress, elbow gloves, white gloves, necklace, earrings</code></pre>Negative, worth keeping:<pre><code>chibi, speech bubble, english text, multiple views, watermark, signature</code></pre><h2 id="version">🧩 Version</h2><ul><li>🔵 v1.0 Anima, for Anima / Qwen-Image DiT checkpoints. ✅ Works in the Civitai on-site generator. Weight 1.0 (usable range 0.85 to 1.0).</li><li>🟠 Remapped Anima 2.9B — this version is the same LoRA remapped onto the 40-layer Anima-2.9B layout (weight 0.5–0.7, no patch node needed). It is a remap, not a retrain, so there may be issues with the 2.9B version — if there is enough interest I may retrain it natively for 2.9B.</li></ul><blockquote>Three triggers, one file. Nothing to swap between outfits. Change the first word.</blockquote><h2 id="tips">💡 Tips</h2><ul><li>The coat has to be asked for. <code>arubaseba</code> alone gives the shirt and the skirt and nothing over them. <code>fur-trimmed coat, coat on shoulders</code> puts it back.</li><li>Never negative her colours. <code>pink hair</code>, <code>horns</code> and <code>halo</code> were stripped from every training caption, which is why they ride the trigger. Negating them fights the LoRA.</li><li>Keep <code>solo</code> in the kimono prompt. Her New Year art is mostly group shots, so without it the rest of the cast turns up.</li><li>Coat colour is a real fork. <code>brown coat</code> and <code>red coat</code> are both trained and land near a coin flip. Name the one you want.</li><li>Kimono knobs: <code>fur collar</code> for the black fur, <code>lace gloves</code> for the hands, <code>single hair bun</code> when the updo comes out loose.</li><li>Gown knobs: <code>coat on shoulders</code> and <code>red coat</code> are trained over the gown. <code>high heels</code> and <code>handbag</code> respond too.</li><li>Alt outfits are easy. The clothes are tags, not the trigger. Dress her in anything and the face holds.</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 ❤️

Rikuhachima Aru 陸八魔アル, 3 Outfits — Blue Archive ブルーアーカイブ [Anima]

🟠 Remapped Anima 3.8B — this version is the same LoRA remapped onto the 52-layer Anima 3.8B / Delta Mix layout (weight 0.65–1.0, plain Load LoRA, no patch node needed). It is a remap, not a retrain, so there may be issues with the 3.8B version.

Files with SHA256: 80465e3f87a51eb5361b6fb810e1b24b3a6e502a961631b4eb727ff53e60e95e