PAI ArtLab模型介绍_人工智能平台 PAI(PAI)-阿里云帮助中心

本文为您介绍Checkpoint（ckpt）、Hypernetwork、embedding、LoRA模型和VAE模型，帮助您初步了解各个模型的特点和效果，从而便捷、高效地使用PAI ArtLab。

Checkpoint（ckpt）

Checkpoint是基础模型或主模型，包括官方的1.5版本和2.1版本，以及用户自定义的特色风格模型。这些模型已完成预训练，可直接用于生成图片。若需个性化调整，可借助其他模型进行微调。

正向prompt：best quality, masterpiece, highres, 1girl,blush,(seductive smile:0.8),star-shaped pupils,china hanfu,hair ornament,necklace, jewelry,Beautiful face,upon_body, tyndall effect,photorealistic, dark studio, rim lighting, two tone lighting,(high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, volumetric lighting, candid, Photograph, high resolution, 4k, 8k, Bokeh
负向prompt：(((simple background))),monochrome ,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, lowres, bad anatomy, bad hands, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, ugly,pregnant,vore,duplicate,morbid,mut ilated,tran nsexual, hermaphrodite,long neck,mutated hands,poorly drawn hands,poorly drawn face,mutation,deformed,blurry,bad anatomy,bad proportions,malformed limbs,extra limbs,cloned face,disfigured,gross proportions, (((missing arms))),(((missing legs))), (((extra arms))),(((extra legs))),pubic hair, plump,bad legs,error legs,username,blurry,bad feet
prompt来源：civitai
效果展示：

Hypernetwork模型需与主模型搭配使用，常用于画风转换和角色生成，但在精细效果上表现一般。

选择模型后，prompt区域会自动生成tag，可通过调整权重影响出图效果。

Embedding模型（又称Textual Inversion模型）需与主模型搭配使用。它可以生成指定角色的特征和风格，体积小，节省空间，但信息量较少，画面和风格还原度不高，使用频率较低。

选择模型后，对应prompt区域会自动生成tag。

正向prompt：best quality, masterpiece, highres, 1girl,blush,(seductive smile:0.8),star-shaped pupils,china hanfu,hair ornament,necklace, jewelry,Beautiful face,upon_body, tyndall effect,photorealistic, dark studio, rim lighting, two tone lighting,(high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, volumetric lighting, candid, Photograph, high resolution, 4k, 8k, Bokeh
负向prompt：(((simple background))),monochrome ,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, lowres, bad anatomy, bad hands, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, ugly,pregnant,vore,duplicate,morbid,mut ilated,tran nsexual, hermaphrodite,long neck,mutated hands,poorly drawn hands,poorly drawn face,mutation,deformed,blurry,bad anatomy,bad proportions,malformed limbs,extra limbs,cloned face,disfigured,gross proportions, (((missing arms))),(((missing legs))), (((extra arms))),(((extra legs))),pubic hair, plump,bad legs,error legs,username,blurry,bad feet
prompt来源：civitai
效果对比（左侧为原图，右侧为模型处理后图片）：

LoRA模型需要搭配主模型使用，无法单独生成图片，LoRA模型是在主模型的基础上训练出来的，所以与搭配的主模型使用效果会更好。在civitai上分享的预训练LoRA模型，作者通常会注明指出该模型是基于哪个主模型进行训练的，用户可以选择对应的主模型与其配合使用，提升生成效果。

单击LoRA模型，即可在prompt中生成权重为1的tag。权重可进行调整（0~1），推荐值为0.8，可根据实际情况调整，值越大越接近LoRA模型。

原模型

LoRA模型

正向prompt： best quality, masterpiece, highres, 1girl,blush,(seductive smile:0.8),star-shaped pupils,china hanfu,hair ornament,necklace, jewelry,Beautiful face,upon_body, tyndall effect,photorealistic, dark studio, rim lighting, two tone lighting,(high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, volumetric lighting, candid, Photograph, high resolution, 4k, 8k, Bokeh
负向prompt：(((simple background))),monochrome ,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, lowres, bad anatomy, bad hands, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, ugly,pregnant,vore,duplicate,morbid,mut ilated,tran nsexual, hermaphrodite,long neck,mutated hands,poorly drawn hands,poorly drawn face,mutation,deformed,blurry,bad anatomy,bad proportions,malformed limbs,extra limbs,cloned face,disfigured,gross proportions, (((missing arms))),(((missing legs))), (((extra arms))),(((extra legs))),pubic hair, plump,bad legs,error legs,username,blurry,bad feet
prompt来源：civitai
随机数种子设置为3，其他参数保持默认，调整LoRA权重，依次查看生成效果。

VAE模型与主模型搭配使用，主要用于调整图片的色彩和滤镜。大部分主模型已内置VAE，无需单独下载，可根据需求选择不同风格的VAE模型。

正向prompt： best quality, masterpiece, highres, 1girl,blush,(seductive smile:0.8),star-shaped pupils,china hanfu,hair ornament, jewelry,Beautiful face,tyndall effect,photorealistic, dark studio, rim lighting,high detailed skin:1.2), 8k uhd, dslr, soft lighting, high quality, volumetric lighting, candid, Photograph, high resolution, 4k, 8k, Bokeh
负向prompt：(((simple background))),monochrome ,lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, lowres, bad anatomy, bad hands, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, ugly,pregnant,vore,duplicate,morbid,mut ilated,tran nsexual, hermaphrodite,long neck,mutated hands,poorly drawn hands,poorly drawn face,mutation,deformed,blurry,bad anatomy,bad proportions,malformed limbs,extra limbs,cloned face,disfigured,gross proportions, (((missing arms))),(((missing legs))), (((extra arms))),(((extra legs))),pubic hair, plump,bad legs,error legs,username,blurry,bad feet
prompt来源：civitai
随机数种子：为55555，其他参数保持默认。
效果图（左侧为原图，右侧为模型处理后图片）：