<h3 id="v2-introduce:-bjbztpxof"><strong>V2 introduce:</strong></h3><h1 id="this-lora-is-based-on-noobaixl_v-pred1.0-z9vpduk7c"><strong><span style="color:rgb(76, 110, 245)">This Lora is based on NoobAIXL_V-Pred1.0   </span></strong></h1><p><span style="color:rgb(255, 255, 255)">这个版本是根据我的喜好细选数据集进行训练的，并改变训练方法，同时还缩小了文件大小。现在它的画风表现为我个人的偏好，所以失去了通过年份标签如"year 2024"等来精确控制画风。不过v2-mini在一些地方的表现会强于v1，所以具体选择v1还是v2请您根据自己的需要。另外v2-mini版本的使用可能会有一些问题，您可以参考我的工作流进行优化，我的工作流镶嵌在例图的元数据中。</span></p><p><span style="color:rgb(255, 255, 255)">This version features a dataset specifically chosen for my preferences, along with changes to the training methodology, resulting in a smaller file size.Its artistic style now caters to my personal preferences. As a result, it has lost the ability to precisely control the style via year tags such as 'year 2024'.v2-mini exhibits superior performance compared to v1 in certain aspects. Therefore, the choice between v1 and v2 should be made according to your individual requirements. Note that v2 might present some challenges; for optimization, consult my workflow, which is included in the metadata of the example images</span></p><p></p><h3 id="v1-introduce:-obhf4fbzt"><strong>V1 introduce:</strong></h3><h1 id="this-lora-is-based-on-noobaixl_v-pred0.65s-8soksaj8r"><strong><span style="color:rgb(76, 110, 245)">This Lora is based on NoobAIXL_V-Pred0.65s</span></strong></h1><h3 id="you-can-adjust-the-visual-style-by-using-the-'year-2015-year-2024'-range-for-some-changes.-gbtcltugb"><strong>You can adjust the visual style by using the '<span style="color:rgb(64, 192, 87)">year 2015-year 2024</span>' range for some changes.</strong></h3><h3 id="compared-to-the-&quot;wlop&quot;-style-in-the-noob0.65v-model-this-version-better-aligns-with-my-personal-taste.-12jdsjll5"><strong>Compared to the "wlop" style in the Noob0.65V model, this version better aligns with my personal taste.</strong></h3><p><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/d483120e-4666-4238-abe7-d515347cc88f/width=525/d483120e-4666-4238-abe7-d515347cc88f.jpeg" /></p><h2 id="this-model-was-trained-purely-out-of-interest-with-the-main-purpose-being-to-generate-images-that-i-prefer.-please-do-not-use-it-for-other-purposes.-ajymg0tov"><strong>This model was trained purely out of interest, with the main purpose being to generate images that I prefer. Please do not use it for other purposes.</strong></h2><p><span style="color:rgb(134, 142, 150)">Another interesting thing is that the Lora training this time tried very magical parameters and datasets, but because they were too outrageous, I won't go into detail</span></p><p><span style="color:rgb(255, 255, 255)">最后要说明的是，关于为什么noob模型中已包含"wlop"画师标签却训练了这个lora模型：大模型中直接使用“wlop”标签难以精确控制为我喜好的画风，并且会导致水印的出现。所以我选择使用lora模型精确控制我想要的画风，并且训练集为我精选的高品质图片不包含任何水印。</span></p><p><span style="color:rgb(255, 255, 255)">As a final note, regarding why I trained this LoRA model even though the 'wlop' artist tag is already included in the base model: using the 'wlop' tag directly in the base model doesn't allow for precise control over the style to match my preferences, and it can also lead to watermarks appearing. Therefore, I chose to use a LoRA model to achieve the exact style I wanted, and the training dataset consists of high-quality, hand-picked images without any watermarks.</span></p>

Files with SHA256: 6c606788d76cf778f7590db2c9dd9f5b098187c1c1f6af67a749f7c2c7de7ac7