The whole workflow is based on img2img and reactor, aiming to perform stylized portrait face-swap of characters. Please refer to my Youtube channel for the specific principles:
You can also find other workflows I shared in GitHUB:
Basic Setting:
Soiurce Image: Throw a portrait into it
SDXL Empty Latent Image: Set the initial sampling size
Eff. Loader SDXL: SDXL basic setting
LoRA Stacker: Setting LoRA as your wish
Advanced Setting:
Sampler For First Samplering: Obvious as the name suggests
Sampler For Img2img Samplering: Obvious as the name suggests
Optional Upscale Ratio: Optional upscale the image to improve the img2img quality
Upscale Model: Upscale Model Used For Final Upscale
Final Upscale Ratio: Ratio For Final Upscale
Image Chromatic Aberration: Chromatic Aberration Effect For Final Image
Global Seed: Obvious as the name suggests
Basic Prompt:
Positive Prompt: Nothing to say, write as you wish
Negative Prompt: Nothing to say, write as you wish
Styler Mixer & Fianl Prompt:
SDXL Prompt Styler: Mix the styles by these three nodes,"none" for no style
SDXL TUPLE:
SDXL TUPLE: No need to do with it at all
ControlNet Model:
ControlNet Model: For this workflow,It is recommended to use the depth models for SDXL
Core:
Many Nodes:In most cases, no adjustment is necessary. In some cases, you may need to fine-tune the parameters of them
Output:
First Reactor With FaceRestore: Just for compare with final image
Final Preview: Just For Preview
Final Image Save: No need to explain
整个流程基于img2img和reactor，旨在进行人物风格化的肖像换脸。 具体原理请参考我的Youtube频道：
您也可以在GitHUB中找到我分享的其它工作流:
Basic Setting:
Soiurce Image: 扔一张肖像进去
SDXL Empty Latent Image: 设置最初的采样画幅大小
Eff. Loader SDXL: 基本的SDXL设置
LoRA Stacker: 按您所需设置LoRA模型
Advanced Setting:
Sampler For First Samplering: 如命名般显而易见
Sampler For Img2img Samplering: 如命名般显而易见
Optional Upscale Ratio: 可选的缩放，旨在提升img2img的画质，并非必须
Upscale Model: 最终放大图片时所用的放大模型
Final Upscale Ratio: 最终放大图片所用的放大比率
Image Chromatic Aberration: 给最终图片加上色散效果
Global Seed: 如命名般显而易见
Basic Prompt:
Positive Prompt: 没什么可说的，按您所需写吧
Negative Prompt: 没什么可说的，按您所需写吧
Styler Mixer & Fianl Prompt:
SDXL Prompt Styler: 用这三个节点混合风格,设定为“none”，既没有风格。
SDXL TUPLE:
SDXL TUPLE: 不需要做任何改动
ControlNet Model:
ControlNet Model: 在这个工作流中，推荐使用各种SDXL的depth算法模型。
Core:
Many Nodes:多数情况下，不用调了吧。某些特定的情况下，您或许需要微调这些节点中的参数。
Output:
First Reactor With FaceRestore: 看名字应该能知道了，只是用来跟最终画面做对比用
Final Preview: 最终画面的预览
Final Image Save: 不用解释了
