赞
踩
webui界面如图:
- 可以在SD界面 -> 模型管理 --> Stable Fussion模型 找到下载链接;
- 在 https://huggingface.co 中搜需要的模型;
- 在C站搜:https://civitai.com/
- 在LiblibAI模型站搜:https://www.liblib.art/
- 在吐司站:https://aigccafe.net/
- 其他
微调模型,主要用来定制人物,轻巧,训练高效
主要用来美化色彩,很多主模型已经自带
嵌入模型,主要用来调教文本理解能力
超网络模型,主要用来定制画风,风格
网页安装按git clone https://github.com/Mikubill/sd-webui-controlnet上面描述做就行
cd sd-webui-aki-v4.4
git clone https://github.com/Mikubill/sd-webui-controlnet.git extensions/sd-webui-controlnet
git lfs install
git clone https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main extensions/sd-webui-controlnet/models
git clone https://huggingface.co/lllyasviel/sd_control_collection/tree/main models/ControlNet
可参照官网文档:https://docs.novelai.net/,主要有权重系数、标签替换、标签轮转
在文生图参数的基础上,图生图主要参数如下:
反向提示词统一为:nsfw,(worst quality, low quality, normal quality:1.3),lowres,bad anatomy,long neck,
下面提示词均为正向提示词。
masterpiece,best quality,1girl,solo,dream lover,full face,front head,front body,sitting,forest,light rays,
masterpiece,best quality,red rose,in vast,next to 1girl,solo,drink water,
masterpiece,best quality,1little girl, solo, holding, 1red balloon
masterpiece,best quality,1girl,solo,sitting,garden,highres
masterpiece,best quality,1girl,dream lover,beautiful face,full face,upright body,
masterpiece,best quality,1girl,dream lover,beautiful face,full face,upright body,younger,
masterpiece,best quality,little girl,front face,wearing a red dress,jump,two feed can see and off the ground,seeing roses in garden,lora:GuoFeng3.2_Lora:1, Negative prompt: nsfw,(worst quality, low quality, normal quality:1.3),lowres,bad anatomy,long neck,badhandv4,
Steps: 20, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 2762403894, Size: 640x640, Model hash: 879db523c3, Model: dreamshaper_8, Clip skip: 2, Lora hashes: “GuoFeng3.2_Lora: 5d35ac960118”, Version: v1.6.0
ControlNet的使用:
提示词为:masterpiece,best quality,little girl,front face,wearing a red dress,jump,two feed can see and off the ground,seeing roses in garden,lora:GuoFeng3.2_Lora:1,
Negative prompt: nsfw,(worst quality, low quality, normal quality:1.3),lowres,bad anatomy,long neck,badhandv4,Steps: 20, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 2762403894, Size: 640x640, Model hash: 879db523c3, Model: dreamshaper_8, Clip skip: 2, ControlNet 0: “Module: openpose_full, Model: control_v11p_sd15_openpose [cab727d4], Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Save Detected Map: True”, Lora hashes: “GuoFeng3.2_Lora: 5d35ac960118”, Version: v1.6.0
图片如下:
XYZ table:在左下角脚本中选择X/Y/Z plot,已X轴类型为sample为例:
提示词:masterpiece,best quality,little girl,front face,wearing a red dress,jump,two feed can see and off the ground,seeing roses in garden,lora:GuoFeng3.2_Lora:1,Negative prompt: nsfw,(worst quality, low quality, normal quality:1.3),lowres,bad anatomy,long neck,badhandv4,
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 2762403894, Size: 640x640, Model hash: 879db523c3, Model: dreamshaper_8, Clip skip: 2, ControlNet 0: “Module: openpose_full, Model: control_v11p_sd15_openpose [cab727d4], Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0, Guidance End: 1, Pixel Perfect: False, Control Mode: Balanced, Save Detected Map: True”, Lora hashes: “GuoFeng3.2_Lora: 5d35ac960118”, Version: v1.6.0
图形如下:
提示词为:masterpiece,best quality,1girl,sole,dream lover,(beautiful face),full face,upright body,big eyes,younger,flower,light,hands down,
Negative prompt: nsfw,(worst quality, low quality, normal quality:1.3),lowres,bad anatomy,long neck,badhandv4,
Steps: 20, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 459248525, Size: 512x512, Model hash: a83e25fe5b, Model: GuoFeng3.4, Clip skip: 2, Version: v1.6.0,图像如下:
点击图片下面的按钮:[发送图像和生成参数到图生图选项卡]
DeepBooru反推翻译成汉语:模糊、模糊背景、模糊前景、景深、运动模糊、照片_(中)、头发装饰品、头发花、1girl、散景、聚焦、cosplay_photo、花、独奏、看着观众、红唇、亚洲、长发、嘴唇、室内、3d;
整体看DeepBooru反推内容是有一些正确性的,Clip反推错误,不知道为什么,待以后研究,错误内容如下:
*** Error interrogating
Traceback (most recent call last):
File "sd-webui-aki-v4.4\modules\interrogate.py", line 192, in interrogate
self.load()
File "sd-webui-aki-v4.4\modules\interrogate.py", line 121, in load
self.blip_model = self.load_blip_model()
File "sd-webui-aki-v4.4\modules\interrogate.py", line 101, in load_blip_model
blip_model = models.blip.blip_decoder(pretrained=files[0], image_size=blip_image_eval_size, vit='base', med_config=os.path.join(paths.paths["BLIP"], "configs", "med_config.json"))
IndexError: list index out of range
缩放有4中形式:
注意:只是缩放的话,要将重绘幅度Denoising strength调整为0,不然会引起画面变化;
提示词: smile,重绘强度:0.5,结果如下:
局部重绘和涂鸦的区别在局部重绘没有颜色选择按钮;
上传原图到局部重绘,将重绘幅度设为0,结果如下:
可以看出,局部重绘并未改变原图内容,但若将重绘幅度设为0.75,结果如下:
注意:
复制当前图像到涂鸦,使用图画笔,可以调整颜色,涂鸦后图像如下:
重绘幅度设置为0.75,蒙版边缘模糊度30,提示提:flower, television,图片如下:
### 3.2.8 上传蒙版(Inpaint upload)
注意:
下面以蒙版将人涂成黑色,其他部分涂成白色,提示词为:(a rose in vase:1.3),television, 蒙版边缘模糊度为30,重绘幅度为0.75,生成图片如下:
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。