<h3 id="ckrf-v1">CKRF V1</h3>基于ChenkinNoob-XL Rectified-Flow 0.3 （<a target="_blank" rel="ugc" href="https://civarchive.com/models/2363696/chenkinnoob-xl-rectified-flow">https://civarchive.com/models/2363696/chenkinnoob-xl-rectified-flow</a>）版本训练并融合，融合后画风更偏向宣纸水墨效果。<h3 id="ckrf-v1">CKRF V1</h3>Trained and fused based on the ChenkinNoob-XL Rectified-Flow 0.3 version (<a target="_blank" rel="ugc" href="https://civarchive.com/models/2363696/chenkinnoob-xl-rectified-flow">https://civarchive.com/models/2363696/chenkinnoob-xl-rectified-flow</a>), the post-fusion style leans more toward a Xuan paper ink painting effect.<h3 id="v3.0">V3.0</h3>使用新参数训练Lora并融合，调整了融合比例，增强生成稳定性，避免2.0版本中大量出现的肢体错乱和手指乱的问题，缺点是整体光影偏暗，脸部阴影太重，2.0的光影更好些，有对风格特化需求的使用者可以同时继续使用2.0。<h3 id="v3.0">V3.0</h3>Trained LoRA with new parameters and fused them, adjusting the fusion ratio to enhance generation stability and mitigate the limb disarray and finger distortion issues prevalent in version 2.0. The drawback is the overall darker lighting and heavier facial shadows compared to version 2.0, which offers better lighting. Users with specific style preferences can continue using version 2.0 alongside this one.<h3 id="v2.0">V2.0</h3>重新用Zero Terminal SNR和alpha1参数训练了Lora并融合，同时调整了融合比例。现在的笔触更细腻，身体比例和对暗场景的适应性更好。缺点是对close-up的响应不好，容易画成全身像和远视角。<h3 id="v2.0">V2.0</h3>Trained and merged Lora with new parameters like Zero Terminal SNR and alpha 1. Fusion proportions were adjusted too. Now the strokes are now more delicate, and body proportions and adaptability to dark scenes are better, but the disadvantage is that the response to close-up scene prompt is not good enough, making it easy to draw them as long shot scene.<h3>导言</h3>这是一个用于生成水墨混合风格图像的Lycoris，经过360张挑选的写意画、工笔画以及中国风数字绘画内容训练，依照个人审美融合而成，训练基底模型是NOOB V-PRED 1.0。<h3>优点</h3><ul><li>通过调整chinese ink style、traditional media、color ink \(medium\)、ink \(medium\)等关键词权重，可以产出不同类型的半写意风格。</li></ul><h3>缺点</h3><ul><li>模型在绘画不同元素时表现的风格有差异，总体表现为一种比较精细的写意风格，但不稳定。</li><li>肢体扭曲有时会比较严重，手脚偏大。</li></ul><h3>使用方法</h3>推荐权重为1，根据示例内不同关键词，或其他有关于水墨画风的关键词进行调整，获得需要的产出。<h3 id="introduction">Introduction</h3>This is a Lycoris used to generate Chinese ink mixed style images. Trained with 360 selected freehand paintings, meticulous brushwork paintings, and Chinese style digital painting content and merged according to personal aesthetics. The training base model is NOOB V-PRED 1.0.<h3 id="advantage">Advantage</h3><ul><li>By adjusting the weights of keywords such as chinese ink style, traditional media, color ink \ (medium \), ink \ (medium \), etc., you can have various semi-freehand style.</li></ul><h3 id="shortcoming">Shortcoming</h3><ul><li>The style of the model varies when painting different elements, and overall it shows a relatively fine freehand style, but it is unstable.</li><li>The limbs maybe severely twisted and the hands and feet are slightly larger.</li></ul><h3 id="usage">Usage</h3>The recommendation weight is 1. Adjust according to different keywords in the example or other keywords related to ink painting style to obtain the desired output.

Mixed ink style-混合水墨风格

Mixed Ink V3.safetensors

Ink mixed style-混合水墨风格 (NOOB V-PRED)

v3.0

CKRF V1

v2.0

v1.0

V3.0使用新参数训练Lora并融合，调整了融合比例，增强生成稳定性，避免2.0版本中大量出现的肢体错乱和手指乱的问题，缺点是整体光影偏暗，脸部阴影太重，2.0的光影更好些，有对风格特化需求的使用者可以同时继续使用2.0。V3.0Trained LoRA with new parameters and fused them, adjusting the fusion ratio to enhance generation stability and mitigate the limb disarray and finger distortion issues prevalent in version 2.0. The drawback is the overall darker lighting and heavier facial shadows compared to version 2.0, which offers better lighting. Users with specific style preferences can continue using version 2.0 alongside this one.V2.0重新用Zero Terminal SNR和alpha1参数训练了Lora并融合，同时调整了融合比例。现在的笔触更细腻，身体比例和对暗场景的适应性更好。缺点是对close-up的响应不好，容易画成全身像和远视角。V2.0Trained and merged Lora with new parameters like Zero Terminal SNR and alpha 1. Fusion proportions were adjusted too. Now the strokes are now more delicate, and body proportions and adaptability to dark scenes are better, but the disadvantage is that the response to close-up scene prompt is not good enough, making it easy to draw them as long shot scene.导言这是一个用于生成水墨混合风格图像的Lycoris，经过360张挑选的写意画、工笔画以及中国风数字绘画内容训练，依照个人审美融合而成，训练基底模型是NOOB V-PRED 1.0。优点<ul><li>通过调整chinese ink style、traditional media、color ink \(medium\)、ink \(medium\)等关键词权重，可以产出不同类型的半写意风格。</li></ul>缺点<ul><li>模型在绘画不同元素时表现的风格有差异，总体表现为一种比较精细的写意风格，但不稳定。</li><li>肢体扭曲有时会比较严重，手脚偏大。</li></ul>使用方法推荐权重为1，根据示例内不同关键词，或其他有关于水墨画风的关键词进行调整，获得需要的产出。IntroductionThis is a Lycoris used to generate Chinese ink mixed style images. Trained with 360 selected freehand paintings, meticulous brushwork paintings, and Chinese style digital painting content and merged according to personal aesthetics. The training base model is NOOB V-PRED 1.0.Advantage<ul><li>By adjusting the weights of keywords such as chinese ink style, traditional media, color ink \ (medium \), ink \ (medium \), etc., you can have various semi-freehand style.</li></ul>Shortcoming<ul><li>The style of the model varies when painting different elements, and overall it shows a relatively fine freehand style, but it is unstable.</li><li>The limbs maybe severely twisted and the hands and feet are slightly larger.</li></ul>UsageThe recommendation weight is 1. Adjust according to different keywords in the example or other keywords related to ink painting style to obtain the desired output.

Files with SHA256: b7d5620f9bd5c9f5877668c41c5a40da6effdd81dda52f60b47f6511f27e3dd4