Update README.md
update
This commit is contained in:
smthemex
@@ -1,40 +1,57 @@
|
||||
# ComfyUI_HiDiffusion_Pro
|
||||
A HiDiffusion node for ComfyUI
|
||||
|
||||
2024-06-01 update
|
||||
-----
|
||||
--修复使用加速模型时，controlnet无效的错误。
|
||||
--DMD2加速模型已可以正常使用
|
||||
--内置图片裁切功能
|
||||
--模型加载节点连接controlnet_repo节点时，可以为空或者其他不支持的模型，此时，text2img 和img2img生效，就是文生图和图生图，没有controlnet功能，也就是说文生图节点其实可以删掉了
|
||||
--菜单加入timestpes，对于LCM模型，可以试着改成例如390，128或者其他小于999的数字，或许效果会更佳。
|
||||
|
||||
Previous updates
|
||||
--增加了SDXL-scribble 模型的支持。
|
||||
--增加了model.yaml文件，你可以修改该文件，添加其他的可能支持“扩散模型”或者“controlnet”或者“unet”模型 ，所有目前支持的模型的repo_id都在该文件的注释里,除了openpose和inpainting
|
||||
--增加了SDXL flash 加速扩散模型的支持，也就是说controlnet节点可以的SDXL模型，还可以用SDXL flash 扩散模型
|
||||
--增加了对anyline（基于mistuline模型）和controlnet 线稿类的canny支持，增加了openposeXL模型的支持（须搭配controlnet openpose）
|
||||
--所有节点都支持XL-lighting\Hyper\LCM\DMD2\ Unet加速，建议适当提高步数
|
||||
|
||||
--Fix the error where controllnet is invalid when using accelerated models.
|
||||
--The DMD2 acceleration model is now functioning properly
|
||||
--Built in image cropping function
|
||||
--When the model loading node is connected to the controllet-repo node, it can be empty or other unsupported models. In this case, text2img and img2img take effect, which are the text generated graph and graph generated graph. There is no controllnet function, which means that the text generated graph node can actually be deleted
|
||||
--Add timestepes to the menu. For the LCM model, you can try changing it to 390 or other numbers smaller than 999 for better results.
|
||||
Previous updates
|
||||
--Added support for SDXL capable models.
|
||||
--Added model.yaml file, you can modify it and add other models that may support "diffusion model", "controllet", or "unet". The repo_id of all currently supported models is in the comments of this file, except for openpose and repainting
|
||||
--Added support for the SDXL flash accelerated diffusion model, which means that the control net node can use the SDXL model, and the SDXL flash diffusion model can also be used
|
||||
--Added Canny support for Anyline (based on the pipeline model) and Controlnet draft classes, and added support for Openpose XL model (to be paired with Controlnet Openpose)
|
||||
--All nodes support XL lighting, Hyper, LCM, DMD2, Unet acceleration, it is recommended to increase the number of steps appropriately
|
||||
|
||||
NOTICE
|
||||
----
|
||||
注意，如果使用本地下载的模型，注意要配套config文件，以及有些模型是需要bin模型，或者模型名称要一样，注意看报错信息提示缺失什么模型。 使用openpose的大模型会预下载VAE模型，无梯子的注意这里会报错。
|
||||
|
||||
HiDiffusion From: [link](https://github.com/megvii-research/HiDiffusion)
|
||||
----
|
||||
|
||||
My ComfyUI node list：
|
||||
-----
|
||||
|
||||
1、ParlerTTS node:[ComfyUI_ParlerTTS](https://github.com/smthemex/ComfyUI_ParlerTTS)
|
||||
|
||||
2、Llama3_8B node:[ComfyUI_Llama3_8B](https://github.com/smthemex/ComfyUI_Llama3_8B)
|
||||
|
||||
3、HiDiffusion node：[ComfyUI_HiDiffusion_Pro](https://github.com/smthemex/ComfyUI_HiDiffusion_Pro)
|
||||
|
||||
4、ID_Animator node： [ComfyUI_ID_Animator](https://github.com/smthemex/ComfyUI_ID_Animator)
|
||||
|
||||
5、StoryDiffusion node：[ComfyUI_StoryDiffusion](https://github.com/smthemex/ComfyUI_StoryDiffusion)
|
||||
|
||||
6、Pops node：[ComfyUI_Pops](https://github.com/smthemex/ComfyUI_Pops)
|
||||
|
||||
7、stable-audio-open-1.0 node ：[ComfyUI_StableAudio_Open](https://github.com/smthemex/ComfyUI_StableAudio_Open)
|
||||
|
||||
8、GLM4 node：[ComfyUI_ChatGLM_API](https://github.com/smthemex/ComfyUI_ChatGLM_API)
|
||||
|
||||
9、CustomNet node：[ComfyUI_CustomNet](https://github.com/smthemex/ComfyUI_CustomNet)
|
||||
|
||||
10、Pipeline_Tool node :[ComfyUI_Pipeline_Tool](https://github.com/smthemex/ComfyUI_Pipeline_Tool)
|
||||
|
||||
11、Pic2Story node :[ComfyUI_Pic2Story](https://github.com/smthemex/ComfyUI_Pic2Story)
|
||||
|
||||
12、PBR_Maker node:[ComfyUI_PBR_Maker](https://github.com/smthemex/ComfyUI_PBR_Maker)
|
||||
|
||||
Notice（节点的特殊功能说明 Special Function Description of Nodes）
|
||||
-----
|
||||
--优化了节点，现在只有采样器和模型加载2个节点，优化了模型加载流程；
|
||||
--支持SDXL-lighting\Hyper\LCM\DMD2\的加速Unet，建议适当提高步数；
|
||||
--基于官方的更新，加入lora支持，需要填关键词；
|
||||
--加入skip，去掉意义不大的其他参数；
|
||||
--支持所有的SDXL controlnet模型（须配置config文件）；
|
||||
--你可以修改model.yaml文件，添加其他的可能支持“扩散模型”或者“controlnet”或者“unet”模型；
|
||||
--所有节点都支持XL-lighting\Hyper\LCM\DMD2\ Unet加速，建议适当提高步数；
|
||||
--支持大部分的基于SDXL的扩散模型.
|
||||
|
||||
--Optimized the nodes, now there are only two nodes: sampler and model loading, and the model loading process has been optimized;
|
||||
--Supports acceleration Unet for SDXL lighting \ Hyper \ LCM \ DMD2 \, it is recommended to increase the number of steps appropriately;
|
||||
--Support Lora, keywords need to be filled in;
|
||||
--Add skip and remove irrelevant parameters;
|
||||
--Supports all SDXL controllnet models (requires configuration of config file);
|
||||
--You can modify the model.yaml file and add other models that may support "diffusion models", "controllnet", or "unet";
|
||||
--All nodes support XL lighting, Hyper, LCM, DMD2, Unet acceleration, it is recommended to increase the number of steps appropriately ;
|
||||
--Supports most SDXL based diffusion models.
|
||||
|
||||
|
||||
1.Installation
|
||||
-----
|
||||
1.1 In the .\ComfyUI \ custom_node directory, run the following:
|
||||
@@ -49,8 +66,7 @@ HiDiffusion From: [link](https://github.com/megvii-research/HiDiffusion)
|
||||
need diffusers >=0.27.0
|
||||
yaml
|
||||
|
||||
|
||||
3 Download the model
|
||||
3 About models
|
||||
----
|
||||
3种调用模型节点的方法
|
||||
一种是,在repo_id填写诸如"stabilityai/stable-diffusion-xl-base-1.0" 这样的标准抱脸的repo_id,系统会自己下载或者查找.cache路径，这个要开全局外网或者你开启了hf镜像
|
||||
@@ -73,23 +89,17 @@ For partially supported models, please refer to the model.yaml file
|
||||
|
||||
5 example
|
||||
-----
|
||||
use openpose XL model
|
||||
|
||||
|
||||
|
||||
use canny XL model or mistoline or scribble（anyline）
|
||||
|
||||
|
||||
|
||||
use controlnet and lora
|
||||
|
||||
|
||||
use inpainting XL model
|
||||
|
||||
|
||||
txt2img use XL/sd1.5/XL turbo/background...
|
||||
|
||||
|
||||
|
||||
img2img use XL XL-flash or base XL
|
||||
|
||||
|
||||
|
||||
6 Citation
|
||||
------
|
||||
|
||||
Reference in New Issue
Block a user