<h2 id="heading-6">V3</h2><p>多分辨率噪声版本。推荐权重0.6，具有更换衣服和NSFW的能力。更换衣服保留头部提示词即可。</p><p>multires version.the recommended weight is 0.6 and have the ability of changing clothes and NSFW.keep head prompt when changing clothes</p><h2 id="heading-325">V2</h2><p>提高了质量。换装降低权重到0.5左右，同时保留头部提示词。默认服装推荐权重0.8以上。</p><p>improved quality.reduce the weight to about 0.5 to chane the clothes,while keeping the head prompts.the recommend weight for default clothes is above 0.8.</p><h3 id="heading-873">中文</h3><p>我根据要求做了这个lora。"kiana"是触发词，"twin braids,long braid,hairband,gloves,black shorts,white legwear"可以获得更加稳定的输出。推荐权重0.8，想要改变服装的话适当降低权重到0.6左右，并且仅保留触发词及头部相关提示词以单独获得她的头。白色类型的上装有概率出现默认服装，需要抽奖。</p><h3 id="english">English</h3><p>I made this lora upon request. "kiana" is the trigger word, and "twin braids, long braid, hairband, gloves, black shorts, white legwear" can get a more stable output. The recommended weight is 0.8. If you want to change the clothing, reduce the weight to about 0.6, and only keep the trigger words and head-related prompt words to obtain her head alone. There is a probability of default clothing appearing on white tops, so you need to try a few more times.</p><h3 id="heading-872">不允许商业用途</h3><h3 id="no-commercial-usage">NO COMMERCIAL USAGE</h3>

Kiana-White Comet-Honkai 3rd

<p>新的数据集，更换了训练模型，提高了输出质量，但仍然存在一些问题。我还在学习中。</p><p>new dataset,changed the training model,improved output quality,still have some questions.still learning.</p>

Files with SHA256: 9f2362ad43ba686a0de42271ef2e9e3dc9d1d25cf66d6d60bd0a7152e8f845d8