# Comfy UI 筆記
### **comfyUI workflow**
* https://drive.google.com/drive/folders/1t1Yz3ZU-qDH7nO8mUrBFCDIEhAbKAUZV?usp=sharing
* https://github.com/602387193c/ComfyUI-wiki/tree/main/pysssss-workflows
# 基礎操作
* https://vocus.cc/article/654798f8fd89780001d4ffaa
* https://www.uisdc.com/stable-diffusion-comfyui-8
# 節點安裝
* Sage Attention:https://www.youtube.com/watch?v=VMT-7ocHYHQ
## 壞掉的節點
* magicanimate
* reactor(疑似python衝突)
* 3Dpack
* ipadapter改為ipadapter advanced
* controlNext
* easy use
* anytext
* kolorswapper
## 待研究
* **Loop** https://www.bilibili.com/video/BV1uN4y197wu/?vd_source=d060fe4dd5a6683127f1df9c93c14b1f
* https://github.com/SeargeDP/SeargeSDXL

# 工具
* https://www.youtube.com/watch?v=WieHllvKZOY
* 工具 https://github.com/ZHO-ZHO-ZHO/ComfyUI-Workflows-ZHO

## 圖生衣服
* https://www.youtube.com/watch?v=xsz2T1hl7gQ
## 文生影片
* 控制臉部:https://openart.ai/workflows/shark_joyful_41/svd-face-control/URVGpGGRiVuQTiRiABWV
* https://github.com/kijai/ComfyUI-HunyuanVideoWrapper
* 安裝教學:https://medium.com/@dseditor/comfyui-%E6%B7%B7%E5%85%83hunyuan-video%E7%9A%84%E9%85%8D%E7%BD%AE%E8%88%87windows%E5%BA%95%E4%B8%8Btrition%E5%AE%89%E8%A3%9D-cad66f32ace3
## 圖生影片
* https://github.com/Lightricks/LTX-Video
* SVD 適合背景:https://blog.csdn.net/2401_84760527/article/details/144077667
## 文生衣服
* https://github.com/tryonlabs/FLUX.1-dev-LoRA-Outfit-Generator
* https://huggingface.co/tryonlabs/FLUX.1-dev-LoRA-Outfit-Generator
## 字體復刻
* anytext
1. 修改文字:https://www.youtube.com/watch?v=fe2DfmqD_C4
* 尺寸最大768 要能被64整除
1. 增加文字:https://www.youtube.com/watch?v=EKbmUMPWWY4
2. https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-wanxiang-api-for-anytext
* Kwai Kolors
3. https://www.youtube.com/watch?v=BpYXy-t0GSU
## 一键生成多视角角色列表
* https://www.youtube.com/watch?v=DEO9FhQv1xc
## 多圖混和控制
* https://www.youtube.com/watch?v=Zv_zGm07qjY
* https://www.youtube.com/watch?v=b2SOf242xkY
## LTX-Video
* https://www.youtube.com/watch?v=94Zt9ByW9Vo
## Lora訓練
* https://www.youtube.com/watch?v=y9tCgzXMpy8
## SV3
https://www.youtube.com/watch?v=hUjvk9u_mTk
## 漫畫生成
* https://www.youtube.com/watch?v=_hKIK6YKd40
## InstantID
* https://github.com/ZHO-ZHO-ZHO/ComfyUI-InstantID
## controlnext

## controlnet
* https://zhuanlan.zhihu.com/p/680978217
* https://www.youtube.com/watch?v=ItSd7SdBeuY
* https://www.bilibili.com/video/BV1gT4m1S78K/
* https://blog.hinablue.me/comfyui-animatediff-vid2vid-workflow/

## 實時繪圖

## 3D
* https://github.com/flowtyone/ComfyUI-Flowty-CRM?tab=readme-ov-file
## step preview
* https://github.com/ltdrdata/ComfyUI-Inspire-Pack
* 

## 轉影片
* SVD:https://www.youtube.com/watch?v=dPTdSoryaCQ
* https://www.patreon.com/posts/95413719
* 較接近於換臉:https://www.youtube.com/watch?v=U0JFVZWf8Is
* 分解影片-生成-組合:https://www.youtube.com/watch?v=PRz-Pg0dad4
* 成果:file://wsl.localhost/Ubuntu/home/ubuntu/ComfyUI/output/ComfyUI_00128_.webp
* densepose跳舞:https://github.com/magic-research/magic-animate

## 穿衣
### 影像換衣
* https://www.bilibili.com/video/BV1S1421f7BP/
confyui無法搭建
* https://www.youtube.com/watch?v=_mYe8wSUaTg
臉+動作+服裝
### 指定服飾
* https://www.youtube.com/watch?v=1dn66kcqdUU
### flux
* https://www.youtube.com/watch?v=NoniP1-b_0g

## Animate Anyone
* 文章說明 https://futureweb.pro/tools/ai-web3-blog/openpose-animate-anyone-ai-model
* 影片教學
* https://www.youtube.com/watch?v=TUFUOhI3B4U
* https://www.youtube.com/watch?v=IQFdNYrNgOc
* 真人影片轉骨架影片網站:https://replicate.com/lucataco/vid2openpose
* 照片+骨架影片啟動程式:https://colab.research.google.com/github/camenduru/Moore-AnimateAnyone-colab/blob/main/Moore_AnimateAnyone_colab.ipynb
## Magic Animate

## segment
* 自動偵測:https://www.youtube.com/watch?v=EFFJDfewbUU
* https://colab.research.google.com/github/camenduru/grounded-segment-anything-colab/blob/main/grounded-segment-anything-colab.ipynb#scrollTo=nsa5xDeNgIGH
* https://github.com/storyicon/comfyui_segment_anything?tab=readme-ov-file
* mask:https://segment-anything.com/demo
* https://www.youtube.com/watch?v=Yj4a1gDfHvI
* https://openart.ai/workflows/piranha_fumbling_24/ipadapter-attention-masking---automatic-selection-using-segment-anything/YfxsTCU5hO5XaQxaHrPz

缺:人物為一體，衣服無法分割-->不適用換衣

## mask換衣
## 跳舞
## inpaint
## swap face
## img2vid
## 修復
* https://www.youtube.com/watch?v=qa8WWHlTvcI&t=446s
* https://www.youtube.com/watch?v=EmDLNRPmdmg
* https://www.youtube.com/watch?v=CxB47DMEyYQ&t=464s
* 手部:https://www.youtube.com/watch?v=yBgDZuj3svk&t=8s

## 批量生圖
* 文生圖
* https://www.douyin.com/search/comfyui%E5%9B%BE%E7%94%9F%E5%9B%BE%E6%89%B9%E9%87%8F%E5%A4%84%E7%90%86?modal_id=7339202908136738100&publish_time=0&sort_type=0&type=general
* https://blog.csdn.net/JuMengXiaoKeTang/article/details/136243673
## 圖轉3D
* https://www.bilibili.com/video/BV1yw4m1o7GB/?spm_id_from=autoNext&vd_source=d060fe4dd5a6683127f1df9c93c14b1f
* https://youtube.com/watch?v=-okTLokYrx8
* **LCM** https://github.com/0xbitches/ComfyUI-LCM?tab=readme-ov-file
* UV圖轉3D:https://blog.csdn.net/jacke121/article/details/118558667
* 拆解UV:https://www.zaowuyun.com/article/6508.html

## 細節&放大
* SD-https://www.youtube.com/watch?v=ev5tpmEzZFE

## 換臉
* https://openart.ai/workflows/jinqingwen0727/faceswap-with-instantidipadapter/lG0ZTDCK2IOK9jLDqM8t

## 其他
* ComfyUI Templates:https://www.patreon.com/posts/mian-fei-xia-zai-96283854
* Comfyui+mixlab，让ai人物随你的骨骼变换做动作:https://www.bilibili.com/video/BV1Ap4y1f7iu/?vd_source=d060fe4dd5a6683127f1df9c93c14b1f
* XY-plot:https://www.youtube.com/watch?v=0tdy9arZJqM

# 節點
* LCM 建議 8-12 step
* prompt
* 自動生成
* 自定義
* B-A圖

* 臉部修復
* 
* 高清
* 放大
* 
* animatediff
* IPAdapter
* 自動偵測遮罩-GroundingDinoSAMSegment (segment anything)
* EasyLoraStack:lora整理
* LayerMask:SegformerB2ClothesUltra:自動遮罩
* Pulid

## 參數
* ksampler
* 參考:https://stable-diffusion-art.com/samplers/
* sampler&scheduler
1. 快速、融合、新穎且品質良好的東西(速度好):
**DPM++ 2M Karras**，20 – 30 個步驟
UniPC 有 20-30 個步驟。
2. 高品質的影像並且不關心收斂(品質優良):
**DPM++ SDE Karras** 有 10-15 個步驟（注意：這是一個較慢的採樣器）
DDIM 有 10-15 個步驟。
3. 穩定、可重複的影像，請避免使用任何祖先取樣器{a開頭}。
4. 簡單的東西:Euler 和 Heun 是不錯的選擇。 減少 Heun 的步驟以節省時間。
* 採樣步數越多，任何兩個步之間的雜訊降低越小。這有助於減少截斷錯誤。