VRChat Kon

Based on the animagineXLV3_v30 training, generate images of the 3D model Kon sold on Booth.

Recommended weight: 0.7 or 0.9 is recommended. The higher the value, the more the image style leans towards rendering, and the more fixed the character's appearance becomes. 0.7 is a more balanced value, resulting in a 2D image style. You can refer to the example image xyz.

The dataset is trained at a resolution of 1152*1152 with 60 images.

Trigger words: 'vrchat kon' for the original character, and 'vrchat kon clothes' for the original clothing. If the default clothing is too difficult to remove, it can be placed in the reverse prompt words.

The labeling method involves removing character features + quality table + style table.

=========================================================

基于animagineXLV3_v30训练，生成Booth上售卖的3D模型Kon图像。

推荐权重：推荐0.7或0.9。数值越高图像风格越偏向渲染，而且形象越固定。0.7是一个比较均衡的数值，会有2D图像风格。可以参考例图的xyz。

数据集以1152*1152分辨率进行训练，有60张。

触发词：vrchat kon，触发原始衣服用 vrchat kon clothes，如果默认衣服太难去除可以放在反向提示词里。

打标方式是删除人物形象特征+质量表+画风表。