Linuxmint21.2 折腾调校笔记【2023年12月更新……】

[yuzh652800]

接上一楼，
默认安装完后，model文件夹下 等文件夹都是空的，需要自己下载模型，放到相应文件夹。

SDV1.5：https://github.com/easydiffusion/sdkit- ... afetensors
GFPAN：https://github.com/TencentARC/GFPGAN/re ... ANv1.4.pth

或者linuxmint下，更建议直接安装easydiffusion，显卡安装好，CUDA安装好，使用下面命令直接安装，基础模型都会自动下载，不用手动配置。

easydiffusion项目地址：https://github.com/easydiffusion/easydiffusion

[hilario]

网易云音乐，楼主是怎么安装运行的，可以分享一下宝贵经验吗？谢谢~

[yuzh652800]

网易云音乐，楼主是怎么安装运行的，可以分享一下宝贵经验吗？谢谢~

安装星火商店 https://spark-app.store/，常用的软件都有，直接在星火商店安装。

[yuzh652800]

引自https://www.bilibili.com/video/BV1894y1v7G6/?sp ... 7e8a2ec641

# 1.克隆项目仓库

代码： 全选

git clone https://github.com/comfyanonymous/ComfyUI

# 2.创建python3虚拟环境,避免环境污染

代码： 全选

virtualenv -p python3 ComfyUI

# 3.进入ComfyUI目录

代码： 全选

cd ComfyUI

# 4.进入python3虚拟环境

代码： 全选

source ./bin/activate

# 5.安装ComfyUI项目依赖
# 因为我用的是NVIDIA显卡

代码： 全选

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118 xformers
pip install -r requirements.txt

# 5.然后就可以运行主程序了

代码： 全选

python main.py

# 6.用软连接与stableDiffusion共享模型
# ComfyUI的模型目录在 /home/jianbo/桌面/ComfyUI/models/checkpoints
# SD 的模型目录在 /home/jianbo/桌面/stable-diffusion-webui/models/Stable-diffusion

# 创建软链接的命令为
# ln -s /path/to/shared_file /path/to/symlink
# 其中，/path/to/shared_file 是共享的原始文件的路径，
# /path/to/symlink 是软链接的路径和名称。

ln -s /home/jianbo/桌面/stable-diffusion-webui/models/Stable-diffusion /home/jianbo/桌面/ComfyUI/models/checkpoints

代码： 全选

ln -s /media/yuzh/ZhiTai1T/aisd/yuzhmodels/lora Lora
ln -s /media/yuzh/ZhiTai1T/aisd/yuzhmodels/stable-diffusion Stable-diffusion

# 到这里就全部完成了,可以看到启动是非常快的



一台电脑装多套Stable Diffusion时共享模型文件的方法
https://www.bilibili.com/read/cv27218597/

—————————————————2024.3.3补充—————————————————————————
ComfyUI : NEW Official ControlNet Models are released! Here is my tutorial on how to use them.
https://www.youtube.com/watch?v=DMxnf4WXMsY

It's official! Stability.ai has now released the first of our official stable diffusion SDXL Control Net models. In this ComfyUI tutorial we will quickly cover how to install them as well as how to use the preprocessors to get things all set for the Control Net node. We will also cover installing the manager, since it makes things a lot easier going forward. I did cover that in Episode 2 in-depth, so if you need a refresher, check out that video as well. Please note, these are now the official models, not the ones released by others.

Want to help support the channel and get this graph? WOOT!


/ @sedetweiler

You can grab the official Control Net LoRA (OFFICIAL) files from here: https://huggingface.co/stabilityai/co... (Remember, these are Control Net files and are LoRA in name only!)

代码： 全选

git lfs install
git clone https://huggingface.co/stabilityai/control-lora

官方对于姿态估计的说明：https://github.com/Fannovel16/comfyui_c ... estimators

Grab the manager custom node from here: https://github.com/ltdrdata/ComfyUI-M...

Installation[method3] (Installation for linux+venv: ComfyUI + ComfyUI-Manager)
To install ComfyUI with ComfyUI-Manager on Linux using a venv environment, you can follow these steps: prerequisite: python-is-python3, python3-venv

Download scripts/install-comfyui-venv-linux.sh into empty install directory
ComfyUI will be installed in the subdirectory of the specified directory, and the directory will contain the generated executable script.
chmod +x install-comfyui-venv-linux.sh
./install-comfyui-venv-linux.sh

SDXL model from here (OFFICIAL): (bonus LoRA also here)
https://huggingface.co/stabilityai/st...


Control Net Preprocessors are available here:
https://github.com/Fannovel16/comfyui...
Remember, you will have to restart the ComfyUI interface for these to start loading, and the models will also download as they are requested from the Controlnet processor.

Twitter:

/ sedetweiler
Threads: https://www.threads.net/sedetweiler

[yuzh652800]

stablediffusion1.6中添加人脸修复选项的方法，在设置>全部显示中，找到

[yuzh652800]

DaVinci Resolve Can’t Import Mp4 Media (EASY Fix! 2023)
https://beginnersapproach.com/davinci-r ... ort-image/

看视频信息用Mediainfo

1.用Handbrake将mp4转换为H.264(x264)，参数设置参照：


2.命令行转换：

代码： 全选

ffmpeg -i h265.mp4 -c:v dnxhd -vf "scale=3840:2160,fps=24000/1001,format=yuv422p10le" -profile:v dnxhr_hqx -c:a pcm_s16le -ar 48000 -hide_banner outputfile.mxf

Things to note in the command line:

-i h265.mp4 = Your input filename with any format. (“-i” should be included)
-c:v dnxhd = FFMPEG encoder DNxHD – used for DNxHR encoding.
-vf = specifies video filter settings as given below.
scale=1920×1080 = your required resolution. You can change it to 3840:2160 for 4K UHD.
fps=24000/1001 = setting 23.976 frame rate.
format=yuv422p10le = Changing to 4:2:2 10-bit from 8 bit. For 8-bit output, use this value “yuv422p”.
-profile:v dnxhr_hqx = Setting the DNxHR HQX codec. For 422 8-bit, use “-profile:v dnxhr_hq”
-c:a pcm_s16le = Convert your audio to 16-bit WAV format.
-ar 48000 = Set audio frame rate to 48KHz.
-hide_banner = this command hides unnecessary build information.
outputfile.mxf = Your output filename .mxf.

3.不支持10bit编码，改用8bit编码
Handbrake = Go to the “Video” tab and choose “H.264 x 264” Video encoder for 8 bit encoding.
FFMPEG = Choose the format and profile as “format=yuv422p” -profile:v dnxhr_hq” for 422 8-bit encoding.

[yuzh652800]

I added setlocation (LC_CTYPE, ""); in line 423 of my local source intern/ghost/intern/GHOST_SystemX11. cc recompile and it will work fine.

bool GHOST_SystemX11::openX11_IM()

{
if (!m_display) {
return false;
}

/* set locale modifiers such as `@im=ibus` specified by XMODIFIERS. */
setlocale(LC_CTYPE,"");//New
XSetLocaleModifiers("");

m_xim = XOpenIM(m_display, nullptr, (char *)GHOST_X11_RES_NAME, (char *)GHOST_X11_RES_CLASS);
if (!m_xim) {
return false;
}

引自：https://projects.blender.org/blender/blender/issues/87578

[yuzh652800]

在linux下使用Unreal虚幻引擎时，点击内容管理器左上角添加quixel内容，总显示failed to restart background service或者background services stopped

经查询，根源是bridge插件需要的nodejs没有运行，Windows下经常是被杀毒软件拦截，linux下经常是没有权限。
这个程序在/opt/Unreal5Linux/Engine/Plugins/Bridge/ThirdParty/Linux下

将这个目录下的node-bifrost赋予读写权限，可执行程序运行权限。
再次运行bridge,就可以了。

参考帖子：https://forums.unrealengine.com/t/quixe ... /238156/21

Linux版Unreal engine下载地址：https://www.unrealengine.com/zh-CN/linux

[yuzh652800]

下载YouTube视频还是推荐使用 youtube-dl 这是个命令行程序，没有图形界面。

Deepin终端里执行：

代码： 全选

sudo apt install youtube-dl

​
来安装。用法如下：

代码： 全选

cd ~/视频
youtube-dl --merge-output-format mp4 -f "bestvideo+bestaudio[ext=m4a]" https://www.youtube.com/watch?v=dQw4w9WgXcQ

​
记得把链接替换成你要下载的视频。

这条命令会下载任何编码的最佳画质+AAC编码的最佳音质、自动合并、转封装成MP4文件。


貌似ytDownloader不错：
https://github.com/aandrew-me/ytDownloader

[hilario]

非常感谢，新年快乐

[yuzh652800]

https://blog.csdn.net/weixin_43912621/a ... /135628542

启动
/usr/share/sangfor/EasyConnect/EasyConnect
1
Gtk-Message: 14:31:27.817: Failed to load module “canberra-gtk-module”
(base) fakerth@fakerth-IdeaCentre-GeekPro-17IRB:~/下载$ /usr/share/sangfor/EasyConnect/EasyConnect
Gtk-Message: 14:31:27.817: Failed to load module "canberra-gtk-module"
段错误 (核心已转储)
1
2
3
首先的想法是apt-get install安装这个module：“canberra-gtk-module”，其实并不是没有安装这个module，导致无法启动的原因是libpango这个包在Ubuntu 22.04版本太高了，需要进行降版本处理。为了防止修改系统库带来的风险，直接将相关的so库文件解压到easyconnect同目录下即可，即/usr/share/sangfor/EasyConnect/下。

下载低版本包
http://kr.archive.ubuntu.com/ubuntu/poo ... _amd64.deb
http://kr.archive.ubuntu.com/ubuntu/poo ... _amd64.deb
http://kr.archive.ubuntu.com/ubuntu/poo ... _amd64.deb

[yuzh652800]

命令行参数

https://docs.blender.org/manual/zh-hans ... er-options