<hr /><h2 id="orv2or-1p45habme">|关于V2|</h2><hr />V2是我自己练的，V1是网站练的。V2容量更小，我自己把训练图重新裁剪成1024x1024了，不过832x1216出图好像也没什么问题，倒立的话832x1216更容易出。V2自我感觉柔韧姿势的成功率变高了，不过也不排除是错觉。V2版本的输出名称把日文去掉了，这回动态提示词插件应该不会报错了。V2对画风的影响更弱。V1和V2没有高低之分。总之有爱自取吧。<hr /><h2 id="oror-p6adm27s3">|前言|</h2><hr />这回不是过激色色了,是画各种拉伸姿势的画师。本身数据集没有色色内容。训练集的尺寸太过于随心了..不过还是有推荐尺寸832x1216或者900x1100。推荐搭配其他画风使用，本LORA作为姿势强化lora,虽然画风影响比较重就是了这边建议如果抽到喜欢的姿势，但是有瑕疵的话，可以使用种子右边的变异随机种子，用0.02左右的强度去随机出想要的画面。<hr /><h2 id="oror-7vk4zq3xx">|特点|</h2><hr />1.可爱的蜡笔画风2.各种类似瑜伽一样的拉伸姿势3.成功率低的直立倒立4.享受腿的乱舞<hr /><h2 id="oror-bk427ouv4">|建议词条|</h2><hr />1.full body,barefoot,split,flexible全身，赤足，分离，灵活性。这四个词条就可以随便看能生成什么姿势了。2.leg up,spread legs,from behind,facing away,tiptoes抬起腿,打开腿，后背视角，面向别处，踮脚。这五个随意搭配就行，也不止这五个，反正只要是关于姿势和腿和脚的随意搭配就行。<hr /><h2 id="oror-snoxdpbk8">|后言|</h2><hr />祝玩得开心....虽然不想提，但我的Buzz最近快炼没了...请告诉我怎么打开不用付费可以捐赠的抢先体验模式...我上一个LORA打开了抢先体验，理想金额是500，但没生效，我不知道是怎么回事....总之喜欢我的lora就请捐赠我Buzz吧！ <s>至少把每个lora的底金赚回来</s><hr /><hr /><hr /><hr />| About V2 |<hr />V2 is my own creation, while V1 was trained on the website.V2 is more compact, as I resized the training images to 1024x1024. Interestingly, it still performs well at 832x1216, especially when generating handstands—it seems to handle them better in this size.I feel like V2 has improved flexibility in posing, but hey, that might just be my imagination.Oh, and I removed the Japanese text from the output names, so the dynamic prompt plugin shouldn't throw any errors now.V2 has a lighter impact on the art style.V1 and V2 aren’t really a matter of better or worse; they're just different versions.Anyway, if you like it, feel free to take it!<hr /><h2 id="orintroductionor-qumql02w5">|Introduction|</h2><hr />This time, it's not overly spicy or lewd—it's all about an artist who draws various stretching poses.The dataset itself doesn’t contain any NSFW content.The training set sizes are rather random...but recommended dimensions are 832x1216 or 900x1100.I suggest using this LORA in combination with other styles, as it serves mainly to strengthen poses. However, be aware that it does have a significant influence on the art style.If you stumble upon a pose you like but it has some flaws, try using the variation seed on the right with a strength around 0.02 to randomly tweak the image to your liking.<hr /><h2 id="orfeaturesor-j07uqaeyt">|Features|</h2><hr /><ol><li>Adorable crayon art style</li><li>Various yoga-like stretching poses</li><li>Low success rate for upright handstands</li><li>Enjoy the leg flurry</li></ol><hr /><h2 id="orsuggested-keywordsor-000th81uq">|Suggested Keywords|</h2><hr /><ol><li>Full body, barefoot, split, flexible</li></ol>These four keywords will give you a variety of poses—just see what you can generate.<ol><li>Leg up, spread legs, from behind, facing away, tiptoes</li></ol>These five can be mixed and matched as you like. They’re not the only options; just play around with any keywords related to poses, legs, and feet.<hr /><h2 id="orclosing-remarksor-tyua8jamq">|Closing Remarks|</h2><hr />Have fun...although I hate to mention it, my Buzz is running low...Please tell me how to unlock the Early Access mode without having to pay. I enabled Early Access for my last LORA, set the goal at 500, but it didn't work, and I have no idea why...Anyway, if you like my LORA, please donate some Buzz! <s>At least help me recoup the base cost for each LORA.</s>

ARTIST | 川口イッサ-Kawaguchi Issa

V2。全图手动裁剪成1024x1024,全图词条调整。画风影响更弱，柔韧姿势成功率提升少许，有爱自取。V2. The entire image is manually cropped to 1024x1024, and the entire image entry is adjusted. The influence of painting style is weaker, and the success rate of flexible postures is slightly improved. Love comes to you.

Files with SHA256: 9c604f46de0e16df04dce1b2dbcac03c21c5741c512653d6b7d9516d9ccfc68d