OeAi - Fine-tuned model based on illustrious XL 1.1

• Slightly tweaked illustrious XL 1.1 to be more flat and light-coloerd

• Even without using any danbooru style prompts, it can still perform well in image generation

• Added danbooru tag set style artist name to replicate the style

• You can use it like NAI and train it as a base model

• Using the danbooru tag set prompts has a better effect than illustrious XL 1.1 (better anime illustration style)

• About image generation capabilities:

  This model works well with high-resolution generated images, so it doesn’t need subsampling (the native resolution is 1536x1536, see my example images for details, they all come with comfyui workflow information)

• set up:

  Sampling steps: 26

  CFG: 7

  Sampler: euler_ancestral

  Scheduling: ddim_uniform

  Resolution: 1536x1536, 1536x2112, 1472x2240, 1664x1984

  (To generate horizontal images, reverse the recommended resolution width and height)

OeAi - 基于illustrious XL 1.1的微调模型

• 将illustrious XL 1.1微调至更偏向平涂浅色型风格

• 即使是不使用任何danbooru风格提示词它对图像的生成能力上也能表现良好

• 添加danbooru标签集风格的画师名来复刻风格

• 你可以将它当作NAI一样使用并作为基本模型进行训练

• 使用danbooru标签集提示词相较illustrious XL 1.1有着更好的效果（更好的动漫插画风格）

• 关于图像生成能力：

  此模型在使用高分辨率生成图像时依旧表现良好，因此它无需二次采样（原始分辨率支持1536x1536，具体可以查看我的示例图片，它们都带有comfyui的工作流信息）

• 参数：

  采样步数：26

  CFG：7

  采样器：euler_ancestral

  调度：ddim_uniform

  分辨率：1536x1536，1536x2112，1472x2240，1664x1984

  （生成横图将推荐分辨率宽高反过来）