独创的三步生图法，目的是解决lora、画师tag等对模型生图能力固化的现象，通过多步生成赋予最大的生成操作空间。
这个工作流逻辑简单易懂，包含lora、放大、面部修复，一站式生图。
默认模型仅供测试使用。您可以使用任何您喜欢的模型。
面部检测模型下载地址Anzhc Face seg 640 v4 y11n.pt · Anzhc/Anzhcs_YOLOs at main
眼部检测模型下载地址Anzhc Eyes -seg-hd.pt · Anzhc/Anzhcs_YOLOs at main
4x-UltraSharpV2_Lite模型下载地址https://huggingface.co/Kim2091/UltraSharpV2/tree/main
线条ControlNet下载地址mistoLine_rank256.safetensors · TheMistoAI/MistoLine at main
This workflow logic is relatively simple, but it includes lora, amplification, and repair.
The default model is only for testing purposes. You can use any model you like.
Face detection model download address Anzhc Face seg 640 v4 y11n.pt · Anzhc/Anzhcs_YOLOs at main
Eye detection model download address Anzhc Eyes -seg-hd.pt · Anzhc/Anzhcs_YOLOs at main
4 x - UltraSharpV2_Lite model download address https://huggingface.co/Kim2091/UltraSharpV2/tree/main
ControlNet Line model download address mistoLine_rank256.safetensors · TheMistoAI/MistoLine at main
Description
增加区域控制，调整面部修复参数，添加真寻引导，更新面部检测加载器v4版本（下载链接https://huggingface.co/Anzhc/Anzhcs_YOLOs/blob/main/Anzhc%20Face%20seg%20640%20v4%20y11n.pt）
Enhanced regional control, adjusted facial restoration parameters, added TrueSeek guidance, and updated the facial detection loader to version v4 (download link: https://huggingface.co/Anzhc/Anzhcs_YOLOs/blob/main/Anzhc%20Face%20seg%20640%20v4%20y11n.pt).
FAQ
Comments (14)
Can create this workflow in ENGLISH too? 🙏😭 I couldn't read them
大佬你好，我这k采样器有一个属性错误：'ModuleList' 对象没有 '1' 属性，是因为我下的插件有问题吗
这个工作流至少多少显存可用？大佬
8g已经应该就没问题吧，除非四个模型位置分别加载四个不同的模型
I love the workflow itself. But i made the mistake of trying to remake it from a scratch as a beginner's exercise (cus why not, seemed simple).
Seem like there is some hidden sauce at the coloring step (and some other places; sth with preview?) because i can't for the love of me recreate this to be as efficient (aka without making take way more time and resource). ;-; I am probably just dumb and missing something obvious (again).
So: Thanks for that workflow!
ps. One thing: Anzhc Eyes model seem to be marked as "not safe" on GitHub. Any alternatives?
When I was using Anzhc Eyes, it was not marked as unsafe. However, it is possible to use any other eye detection model.
For your issue, you can try to lower the method steps and increase the magnification rate.
@EKLL Tbh, it seems like upscaling (i assume that's what you mean?) seemed to be skipped on the default (your workflow), since the image size remains the same (unlike in my workflow). But whatever, im using the original for now.
Thanks for a quick reply.
@justdamnedthings545 If you encounter node issues, it’s usually the “efficiency-nodes-comfyui” node that’s causing the problem. Since it has switched branches, you’ll need to right-click and refresh the node several times.