<h2 id="yamabuki-saaya-f.1-dev-lora-model-fhzrsdeb0">Yamabuki Saaya F.1 Dev LoRA Model</h2><p>A LoRA Model based on the base model <strong>FLUX.1 Dev.</strong></p><p>I only use the transparent images for training. So the background will be single-color, not flexible, I think.</p><h2 id="the-usage-0f4tdr8so">The Usage</h2><p>Well, The LoRA can be triggered by the single word 'hanazono tae', However, I still recommend u to add '1girl, solo, blue eyes, long hair, brown hair' after the trigger word. (If u use nunchaku workflows, u even do not need the so-called trigger word)</p><p><s>Also, I only tested the case that the weight of the LoRA is 1.25, and I don't really know if lower value can still make the output normal. (You can test it yourself and tell me in the comment area)</s></p><p>In my tests with the help of <a target="_blank" rel="ugc" href="https://github.com/mit-han-lab/nunchaku">svdquant</a></p><p>I can use the LoRA with the weight of 1. And don't need to add the LoRA file name in the prompt input.</p><h2 id="the-introduction-of-the-character-s2xk8q4ge">The Introduction of the Character</h2><p><a target="_blank" rel="ugc" href="https://danbooru.donmai.us/posts?tags=yamabuki_saaya">Danbooru Posts</a></p><p><a target="_blank" rel="ugc" href="https://en.wikipedia.org/wiki/List_of_BanG_Dream!_characters">Wikipedia</a></p><hr /><p>距离上次发布LoRA时隔<strong>又双叒叕</strong>接近六兆年了（bushi）。（原因懂得都懂，学生公寓的供电情况非常不良，时不时因为所谓的“功率过高违规”而反复断电）</p><h2 id="-28hxtp80u">一些基本的训练参数</h2><p>素材以人物卡面为主，并且由于我存储的卡面都是透明图像，所以最终的输出的背景基本是单色的（也未必，但是在不刻意强调background肯定是不定的background，对于背景的prompt大家可以自行探索），方便后期抠图（确信</p><p>训练脚本是直接采用sd-scripts的sd3分支，并受到fluxgym写script的启发，自行手搓脚本和数据设置而成。</p><p>考虑到本人笔电配置不佳，所以每次训练的效率不高，并且迄今为止找不到办法有效降低avr_loss的办法 <s>(除了拉高steps，但那样更费时间)</s></p><h2 id="lora-pviib3pdx">怎么用这个LoRA</h2><p>跟正常使用SD 1.5的LoRA一样，用FLUX.1 Dev的底模，然后使用触发词'yamabuki saaya'触发LoRA（小提示：如果你采用了跟我一样的nunchaku工作流去加载这个LoRA的话，甚至连这个所谓的触发词都不用写，直接开始写该角色的主要特征），<s>目前测试的权重只有1.25，其它更低或更高的值没有测试。</s>可以使权重为1。当然，为了保证角色特征是正确的，推荐加上"1girl,solo,blue eyes,long hair,brown hair"等该角色的基本特征词，并且尽可能只使用DeepBooru词库的词语。</p><h2 id="-ycbn7zuke">面临的挑战</h2><p>这个LoRA由于只是对这个新角色的训练，因此在肢体等一些细节上仍有很大的不足。遗憾的是，f.1并不像sd 1.5那样可以随意加入embeddings来暂时克服肢体问题。我的评价是，反复抽卡吧(</p><p>（当然如果你有很好的解决方案可以在评论区指点一下，本人感激不尽x）</p>

(BanG Dream!) Yamabuki Saaya F.1 Dev LoRA Model

<p>Nothing here, just a LoRA for a single Anime Character......</p>

Files with SHA256: bb347324976709063296856a2ccaf70a0041e5b1d078dbd832708f690da27971