(日本語説明は後半にあります)<h2 id="about-this-model">About this model</h2>"LilyBreed" is an Illustrious-based SDXL checkpoint, pursuing "kawaii", cuteness. This checkpoint has 2.5D semi real style trained from <a target="_blank" rel="ugc" href="https://civarchive.com/models/783356/pansybreed">PansyBreed</a> outputs.There are 2 variations :<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/955973?modelVersionId=2198552">Basic body style</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/955973?modelVersionId=2198551">"Slender" : more slender body style</a></li><li>To download each types, switch version tabs ↑ above.</li></ul>I feel that model merging is similar to the breeding of flowers (to improve their appearance), so named "Breed".These models have special features :<ul><li>Include many originally trained data to adjust painting style (saturation, tone), eye shape, and quality (finely detailed pupils, finely detailed outlines, fine painting, reducing grain noise etc).</li><li>Detailed backgrounds.</li><li>Powerful prompting based by Illustrious (and NoobXL).</li></ul>More various other LoRA's and Checkpoints are released on my <a target="_blank" rel="ugc" href="https://civarchive.com/user/JujoHotaru/models">CivitAI models list</a> or <a target="_blank" rel="ugc" href="https://huggingface.co/JujoHotaru/lora">HuggingFace repository</a>. Please check it out.<hr /><h2 id="recommended-output-settings">Recommended output settings</h2><ul><li>CFG Scale : 4-5</li><li>Sampler : Eular A</li><li>Sampling steps : 20 - 40</li><li>Resolution : mainly tested on 768x768, 768x1152, 832x1216, 800x1200 and 1024x768</li><li>Using A-Detailer (for face, breasts etc.) is highly recommended. (inpaint denoising strength : 0.3-0.5)</li><li>HiRes Fix : use SwinIR_4x, R-ESRGAN 4x+ or 4x-UltraSharp with denoise strength 0.3-0.7.</li><li>Quality prompt / Negative prompt : See below.</li><li>VAE : embedded.</li></ul><hr /><h2 id="quality-prompts-sample">Quality prompts sample</h2>This and base checkpoints are derived from Illustrious, so use quality prompts for Illustrious models.<pre><code>masterpiece, best quality, amazing quality, very aesthetic, absurdres,
cute face, slender, detailed eyes, detailed hair,
(PUT YOUR PROMPTS HERE)</code></pre><h2 id="negative-prompts-sample">Negative prompts sample</h2><pre><code>lowres, bad quality, worst quality, bad anatomy, sketch, jpeg artifacts, ugly, poorly drawn, blurry, watermark,
bad eyes, bright pupils, white pupils, bad hands, bad face,
(PUT YOUR NEGATIVE PROMPTS HERE)</code></pre><hr /><h2>このデータモデルについて</h2>[ v4.0 アップデート：塗りの質感、背景品質の改善を行いました ]LilyBreedは、二次元イラストレーション生成に特化した「Illustrious」派生形モデルで、<a target="_blank" rel="ugc" href="https://civarchive.com/models/783356/pansybreed">PansyBreed</a>の出力を学習した2.5Dのセミリアル系スタイルとなっており、可愛さを追求しています。以下2種類のバリエーションを用意しました：<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/955973?modelVersionId=2198552">基本頭身タイプ</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/955973?modelVersionId=2198551">Slender : 若干頭身が高めのボディースタイル</a></li><li>各タイプをダウンロードするには、ページ上部のバージョンタブを切り替えてください。</li></ul>私はモデルのマージはお花の交配（品種改良）と近いものを感じているため、品種改良を意味するBreedと名付けました。Breedシリーズの特徴として、作者が独自に学習させた多数のLoRAをマージすることで、画質・色味・陰影や眼の形などを調整したり、輪郭線および眼の質感向上や粒子ノイズ低減、HiRes Fixの高品質化などを図っています。<h3>使い方</h3><ul><li>推奨される出力設定や品質指定プロンプト・ネガティブプロンプトなどは上記を参照してください。</li><li>サンプラーに「Eular A」系統以外を使うと画質が劣化しやすいので注意してください。</li><li>ネガティブプロンプトは空にせず、低品質系のワードを入れて調整したほうが良い画質が出やすいようです。（高速出力などのため使えない場合は無くても大丈夫です）</li><li>A-Detailerの使用を推奨します。顔のリファインのInpaint denoising strengthは0.3、裸体の胸には0.5程度がよいでしょう。</li><li>Illustriousベースのため、PonyベースのBreedシリーズモデルに比べ、当方公開の各種LoRA（のうち、Illustrious対応を銘打っていないもの）の効きが弱くなりますのでご了承ください。</li><li>その他、Illustrious系特有の癖などがありますので、Illustriousモデルの使い方解説記事などを参照してください。</li><li>ネコミミなど動物耳キャラクターを生成する場合、人間耳を出さないようにする<a target="_blank" rel="ugc" href="https://civarchive.com/models/1505391/human-ear-eraser-for-animal-ear">人耳消去LoRA</a>も用意しています。</li></ul>VAEは焼き込み済みですので、別途用意する必要はありません。他にもさまざまなLoRAやデータモデルを公開していますので、<a target="_blank" rel="ugc" href="https://civarchive.com/user/JujoHotaru/models">CivitAIの配布モデルリスト</a>または<a target="_blank" rel="ugc" href="https://huggingface.co/JujoHotaru">HuggingFaceリポジトリ</a>をご覧ください。<hr /><h2 id="merge-recipe">Merge recipe</h2><ul><li>Based on 8 models (all merged with ratio 0.125) :<ul><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/894218">copycatNoob_v10</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/869526">illuminaXLTypemModified_11</a></li><li>illuminous_v01</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/880866">illustPossible_v10</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/827460">librasillustriousxl_v40</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/820208">obsessionIllustrious_v30</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/853496">simpleIllustrious_valt</a></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/910006">zukiILL_v10</a></li></ul></li><li>Original finetuned data with "<a target="_blank" rel="ugc" href="https://civarchive.com/models/783356/pansybreed">PansyBreed v2.0</a>" outputs</li><li>Original trained LoRAs for eye shape/painting/detailing</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/152040">XL_VAE_C F2</a> VAE</li></ul>

LilyBreed

lilybreed_SlenderV50.safetensors

"Slender" v5.0

v6.0

"Slender" v6.0

v5.0

"Slender" v4.0

v4.0

"Slender" v3.0

v3.0

v2.0

v1.0

Created by JujoHotaru"LilyBreed" is an Illustrious-based SDXL checkpoint, pursuing "kawaii", cuteness. This checkpoint has 2.5D semi real style trained from <a target="_blank" rel="ugc" href="https://civitai.com/models/783356/pansybreed">PansyBreed</a> outputs.I feel that model merging is similar to the breeding of flowers (to improve their appearance), so named "Breed".These models have special features :<ul><li>Include many originally trained data to adjust painting style (saturation, tone), eye shape, and quality (finely detailed pupils, finely detailed outlines, fine painting, reducing grain noise etc).</li><li>Detailed backgrounds.</li><li>Powerful prompting based by Illustrious (and NoobXL).</li></ul>More various other LoRA's and Checkpoints are released on my <a target="_blank" rel="ugc" href="https://civitai.com/user/JujoHotaru/models">CivitAI models list</a> or <a target="_blank" rel="ugc" href="https://huggingface.co/JujoHotaru/lora">HuggingFace repository</a>. Please check it out.Recommended output settings<ul><li>CFG Scale : 4-5</li><li>Sampler : Eular A</li><li>Sampling steps : 20 - 40</li><li>Resolution : mainly tested on 768x768, 768x1152, 832x1216, 800x1200 and 1024x768</li><li>Using A-Detailer (for face etc.) is highly recommended. (inpaint denoising strength : 0.3-0.5)</li><li>HiRes Fix : use SwinIR_4x, R-ESRGAN 4x+ or 4x-UltraSharp with denoise strength 0.3-0.7.</li><li>Quality prompt / Negative prompt : See below.</li><li>VAE : embedded.</li></ul>Quality prompts sampleThis and base checkpoints are derived from Illustrious, so use quality prompts for Illustrious models.masterpiece, best quality, amazing quality, very aesthetic, absurdres, cute, kawaii, cute round face, slender, detailed eyes, detailed hair, (PUT YOUR PROMPTS HERE)Negative prompts samplelowres, bad quality, worst quality, bad anatomy, sketch, jpeg artifacts, ugly, poorly drawn, blurry, watermark, bad eyes, bright pupils, white pupils, ugly, bad hands, bad face, (PUT YOUR NEGATIVE PROMPTS HERE)

v5.0 "Slender"

Files with SHA256: fa075011d6946dd7f3db188aee5cee86092975a41eada9d3b9c6599bdacb8db0