新手入門
https://space.bilibili.com/1814756990/channel/collectiondetail?sid=2187799&spm_id_from=333.788.0.0

資料合集
https://github.com/yolain/ComfyUI-Yolain-Workflows
ComfyUI WIKI 速查手冊
Examples of ComfyUI workflows github
Stable Diffusion 基本知識(shí)概論
金子的知識(shí)星球-AI繪畫
SEAART AI科普

一鍵服務(wù)器部署和演示
https://github.com/ComfyWorkflows/ComfyUI-Launcher

web部署和演示
https://blog.csdn.net/gitblog_00723/article/details/141838802

在第一次運(yùn)行StableDiffussion的webui.bat時(shí)，在installing requirements這里停留了半小時(shí)無反應(yīng)，就把cmd關(guān)掉了。再打開webui.bat, 未跳出http地址或者下載庫。最后通過運(yùn)行 python D:\SD\sd-webui-aki-v4.8\sd-webui-aki-v4.8\webui.py，繼續(xù)進(jìn)入下載依賴庫的流程。
然后遇到"No module 'xformers'. Proceeding without it. "的報(bào)錯(cuò)，是通過在webui-user.bat里設(shè)置'set COMMANDLINE_ARGS=--xformers'解決

官方ComfyUI的V0.2.7版本的包使用xformers 的方法:
https://www.reddit.com/r/comfyui/comments/1gnl352/comfyui_starting_log_errors_please_help/
即在用包內(nèi)置的python.exe運(yùn)行以下命令:

D:\ComfyUI_windows_portable_nvidia\ComfyUI_windows_portable\python_embeded>python.exe -s -m pip install -U xformers --index-url https://download.pytorch.org/whl/cu124
Looking in indexes: https://download.pytorch.org/whl/cu124
Collecting xformers
  Downloading https://download.pytorch.org/whl/cu124/xformers-0.0.28.post3-cp312-cp312-win_amd64.whl (168.8 MB)
     ---------------------------------------- 168.8/168.8 MB 978.2 kB/s eta 0:00:00
Requirement already satisfied: numpy in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from xformers) (2.1.3)
Requirement already satisfied: torch==2.5.1 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from xformers) (2.5.1+cu124)
Requirement already satisfied: filelock in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (3.16.1)
Requirement already satisfied: typing-extensions>=4.8.0 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (4.12.2)
Requirement already satisfied: networkx in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (3.4.2)
Requirement already satisfied: jinja2 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (3.1.4)
Requirement already satisfied: fsspec in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (2024.10.0)
Requirement already satisfied: setuptools in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (75.3.0)
Requirement already satisfied: sympy==1.13.1 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from torch==2.5.1->xformers) (1.13.1)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from sympy==1.13.1->torch==2.5.1->xformers) (1.3.0)
Requirement already satisfied: MarkupSafe>=2.0 in d:\comfyui_windows_portable_nvidia\comfyui_windows_portable\python_embeded\lib\site-packages (from jinja2->torch==2.5.1->xformers) (3.0.2)
Installing collected packages: xformers
Successfully installed xformers-0.0.28.post3

提示詞抄作業(yè)網(wǎng)站:
https://openart.ai/(可在線文生圖,圖轉(zhuǎn)prompt，甚至訓(xùn)練Lora模型)
https://civitai.com/
https://arthub.ai/(需要非IE瀏覽器)
Lexica Aperture（可在線文生圖，以及根據(jù)關(guān)鍵字搜圖）
Midjourney提示詞生成器 - 一個(gè)工具箱 - 好用的在線工具都在這里！

DiffMorpher的ComfyUI流程和論文:
https://github.com/AIFSH/DiffMorpher-ComfyUI
https://replicate.com/cjwbw/diffmorpher
https://kevin-thu.github.io/DiffMorpher_page/

• 個(gè)人理解
  按我的理解，
  文生圖的話，基本流程是: 圖像先通過CLIP(Contrastive Language-Image Pre-training 對(duì)比語言-圖像預(yù)訓(xùn)練）把文字編碼成潛在空間，然后傳給Sampler。Sample根據(jù)模型Model和隨機(jī)生成的噪聲，對(duì)此潛空間推理，得到latent格式的圖像，然后通過VAE Decoder解碼成常規(guī)圖片格式的文件
  即如下圖所示:

  
  
  文生圖.png

圖生圖的話，基本流程是在文生圖的基礎(chǔ)上，再加入Load image和VAE Encode兩個(gè)節(jié)點(diǎn)。即如下圖所示

圖生圖.png

術(shù)語解釋:
CFG：表示對(duì)提示詞的忠誠度。值越大，輸出的圖像越接近提示詞的描述，但可能會(huì)降低圖像質(zhì)量，顯得不真實(shí)或者不自然。0表示忽略提示詞。一般在6~8之間。模型也會(huì)有推薦值，比如量化表的Flux 模型就是1.0
降噪(denoise): 表示去噪過程中不保留原圖的程度 1.0 表示完全重新生成; 較低的值會(huì)保留更多原始圖像的特征，太低的話會(huì)導(dǎo)致畫面模糊。圖生圖建議值為0.5-0.7。文生圖的話，填1就可以了。