This is the LoRA of Amakusa Hina (Securities) from "A-Train: All Aboard! Tourism", working best with<a target="_blank" rel="ugc" href="https://civarchive.com/models/260267/animagine-xl-v31"> AnimagineXL V3.1</a>, being used <a target="_blank" rel="ugc" href="https://civarchive.com/models/405165/genimagine-xlfor-lora-training">Genimagine XL(for LoRA training)</a>.It may work with the models mixed with AnimagineXL V3.1.You can also play with some art styles, see below.It seems that some elements of the generation is sometime unstable (the available number of images is probably not enough).<ul><li>The mismatched number of buttons and buttonholes of blue suit jacket</li><li>The incorrect number of buttons of clothes compared with the official [The supposed numbers] Black long dress: 4 Blue suit jacket: 2</li><li>Stuck black hair flower</li></ul>You can try to change the number of sampling steps or the weight of the each token or the LoRA itself, or try to changing the number of the seed.FYI, this LoRA may help to generate some kind of the stock-related images: <a target="_blank" rel="ugc" href="https://civarchive.com/models/122700/pe-stonks-down-meme">https://civarchive.com/models/122700/pe-stonks-down-meme</a><a target="_blank" rel="ugc" href="https://civarchive.com/models/260267/animagine-xl-v31">AnimagineXL V3.1</a>向けの『A列車で行こう はじまる観光計画』の証券社員 天草雛のLoRAです。<a target="_blank" rel="ugc" href="https://civarchive.com/models/405165/genimagine-xlfor-lora-training">Genimagine XL(for LoRA training)</a>を使って学習しました。AnimagineXL_V3.1系のmixモデルでも使えるかもしれません。いくつかの画風を試せます。下を参照ください。いくつかの要素の生成が不安定です。入手できた教師画像の数が足りなかったかもしれません。<ul><li>青いスーツジャケットのボタンとボタン穴の数</li><li>公式と異なるボタン数 [公式のボタン数] 黒いロングドレス: 4 青いスーツジャケット: 2</li><li>取り外しにくい花飾り</li></ul>サンプリングステップ数やLoRA/トークンの強度を変更したり、シード値自体を変更してみたりしてください。一応情報として、このLoRAで株のあるタイプの画像が生成しやすくなるかもしれません。<a target="_blank" rel="ugc" href="https://civarchive.com/models/122700/pe-stonks-down-meme">https://civarchive.com/models/122700/pe-stonks-down-meme</a><h2 id="recomended-settings-sj16bw04f">Recomended settings / 推奨設定</h2>The LoRA weight is around 0.7 to 1.0.The settings below is taken from the recommended settings for AnimagineXL V3.1<ul><li>CFG Scale: around 5-7</li><li>sampling steps: below 30</li><li>sampler: Euler a</li><li>positive prompts for generating higher quality outcomes<ul><li>masterpiece, best quality, very aesthetic, absurdres</li></ul></li><li>negative prompts for generating high-aesthetic images<ul><li>nsfw, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract]</li></ul></li></ul>LoRAのウェイトは0.7から1.0くらいがいいと思います。以下はAnimagineXL V3.1の推奨設定に準拠します。<ul><li>CFG Scale: 5-7 くらい</li><li>sampling steps: 30 以下</li><li>sampler: Euler a</li><li>positive prompts (より高品質な出力結果)<ul><li>masterpiece, best quality, very aesthetic, absurdres</li></ul></li><li>negative prompts (美的な画像)<ul><li>nsfw, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, [abstract]</li></ul></li></ul><h2 id="trigger-words-viozdaoqr">Trigger Words / トリガーワード</h2><h3 id="art-style-x5qtojyxp">Art Style / 画風</h3><ul><li>ATGstyle<ul><li>the 1st and 2nd images / 1枚目と2枚目の画像</li></ul></li></ul>Trigger words of the art styles below are intended to separate the art styles from the entire generation, but you can play with them. 以下のトリガーワードは画風を全体の生成から分離する目的で使用しましたが、試すことはできます。<ul><li>atowpmstyle<ul><li>matured Hina (?) style / 成長した雛 (?)スタイル</li><li>the 5th image / 5枚目の画像</li></ul></li><li>shbrstyle<ul><li>shine and bright style / 光あふれるスタイル</li></ul></li><li>atowpcmstyle</li></ul>[Chibi Styles]<ul><li>atochstyle</li><li>ato3anvcstyle</li><li>atowswcstyle</li></ul>(Unstable / 不安定)<ul><li>bldshystyle<ul><li>strong shadow but shiny style / 濃い影だけどキラキラスタイル</li></ul></li></ul><h3 id="amakusa-hina-7gov02fbv">Amakusa Hina / 天草雛</h3>AmakusaHina, brown eyes, brown hair, twintails, blunt bangsAdd (If necessary) / 必要なら追加<ul><li>dot nose, blush, sidelocks</li></ul>For her facial expressions / 顔の表情用<ul><li>[eyes] ^_^</li><li>[mouth] :3 / peanut mouth</li></ul>[Hair Ornaments and earrings / 髪飾りとイヤリング]<ul><li>black hair flower<ul><li>The hair ornaments may be stack strongly to the image generation, even if without the word. If you want to take this out from the image, try to put this trigger word to the negative prompts.</li><li>この花飾りはトリガーワードに追加しなくても画像生成に強く結びついてしまっているかもしれません。外したい場合は、このトリガーワードをネガティブプロンプトに入れてみたりしてください。</li></ul></li><li>floating earring<ul><li>double diamond-shaped earrings</li><li>二つのひし形イヤリング</li></ul></li></ul><h3 id="clothes-v4wi9b5xz">Clothes / 服装</h3>HinaOutfit, blue suit jacket, open jacket, black long dress, pleated dress, name tagAdd (If necessary) / 必要なら追加buttons, buttonholesUnfortunately, generation of these elements are not stable.<ul><li>The mismatched number of buttons and buttonholes of blue suit jacket</li><li>The incorrect number of buttons of clothes compared with the official [The supposed numbers] Black long dress: 4 Blue suit jacket: 2</li></ul>残念ながら以下の要素の生成は不安定です。<ul><li>青いスーツジャケットのボタンとボタン穴の数</li><li>公式と異なるボタン数 [公式のボタン数] 黒いロングドレス: 4 青いスーツジャケット: 2</li></ul>

Files with SHA256: 75c345520aa964f579a7219f23aef4993493195081e31c75945ffb82b5017856