The workflow has been updated to be fully compatible with the latest version of ComfyUI by replacing several nodes with their newer alternatives.
couple_workflow1.1 in the archive is the older release, while couple_workflow2 is the latest one.
The latest version also includes additional notes for each node.
How to fix the "No module named 'pandas'" error in the latest version of ComfyUI
If clicking “TryFix” in the ComfyUI Manager doesn’t solve the issue,
or
If you get the errorModuleNotFoundError: No module named 'pandas'
Follow these manual installation steps when loading acgem156-ComfyUI:
Go to your ComfyUI folder → open
ComfyUI_windows_portable/python_embeded.Open a Command Prompt in that folder.
Run this command:
.\python.exe -m pip install pandasRestart ComfyUI.
After that, the error will disappear and the node should work normally.
///
For non-portable (regular) ComfyUI installations
If you’re not using the portable version, just install pandas in your current Python environment.
Open a terminal or command prompt and run:
pip install pandasThen restart ComfyUI.
That’s all — the “No module named 'pandas'” error should be gone.
///
This workflow aims to generate multiple characters in a single image using ComfyCouple.
A brief explanation is provided below, but if anything is unclear, feel free to ask in the comments.
https://civarchive.com/articles/11788/multiple-character-for-comfyui-couple
I believe it is usable with almost all checkpoints.
required nodes
Description
FAQ
Comments (30)
能大概說一下怎麼用嗎，譬如prompt是否能雙人nsfw，謝謝~
thanks for sharing!
by it appear Missing Node Types
When loading the graph, the following node types were not found AttentionCouple|cgem156
Sorry, the link seems to be out of date.
https://github.com/laksjdjf/cgem156-ComfyUI
It seems to have been merged here now
@kkr3 dont working This node doesn't appear?
@TheNevyGamer30 The node may not have been installed correctly. Please try running .\python.exe -m pip install pandas following the steps in the description.
@kkr3 So that was it after all.
I thought “pandas” would be something else unrelated.
But now it works :3
作为AI新手，向大佬请教一下：如果两人物叠加面积较大（例如骑单车时，A从B身后环抱，只露出头-手-脚），工作流也能生效吗？画遮罩时，不同角色是不是还需要分层、加标注什么的（类似PS分图层再叠加）？
This is effective even when there are overlapping body positions, but it takes some getting used to.
In particular, if you include prompts for clothing and physical features, the AI will try to describe the clothing as much as possible, which complicates the composition.
Prompts that include character traits are best kept completely separate
@kkr3 大佬，我尝试参考着你的工作流，然后自己搭建工作流。输入的遮罩图片都是用PS画好（设定为1920x1080），再使用“加载图像遮罩”导入。因为是新手，我看不懂你工作流里是怎样设定每次生成的图片数量和指定尺寸，我用的是“空latent”指定1920x1080（为了与遮罩图片的尺寸保持一致）和“K采样器”，结果运行时报错：shape '[4, 8160, 1]' is invalid for input of size 1920。请问怎样解决？
@gallopfeel263
I cannot determine the exact solution because I don't fully understand the situation in which the error occurs, but it is possible that the mask area has abnormal values.
It would be better not to use images created manually in PS or similar software for the mask. Instead, you should use ComfyUI's Mask Editor directly on the generated images.
In my workflow, the "width" and "height" settings inside the Efficient Loader's "Empty latent" correspond to the size of the generated images.
The number of generated images is controlled by the "Batch count" setting of the XY node placed above the Image Loader, as I use XY Input.
Thank you for your reply and patient explanation. After listening to your suggestions and trying multiple times, I finally understand how to use this workflow.
As I become more familiar with this workflow, I increasingly feel that the construction of the entire process is great! Thank you again for sharing!
comfy ui says cr latent switch is from an outdated github page, also xy workflow is where?
Depending on the environment, XYnode may not be able to load, but the cause is unknown.
Please try the 1.1 version attached to this page
https://civitai.com/articles/11788/multiple-character-for-comfyui-couple
CR Latent... is a node included in this
@kkr3 Thank you, I tried using the workflow but it is still very confusing. I am unsure which mask corresponds to what and whether to keep the first mask node place bypassed. Additionally I am unsure where lora 1 and lora 2 and lora 3 etc correspond to as my images only show 1 lora. Is it possible to do a more indepth tutorial
@GreenGoblincel I want to rewrite it, but I don't have the time, so it's impossible.
The first Load image, which can be bypassed, is used for i 2 i and is not normally used for masking.
Lora is applied to the entire range, so you cannot choose where to apply it. Everything read by LoraStacker on the right side of the K sampler is applied.
how can we create overlapped characters as when the region overlapped the character got weird
For complex compositions, it’s difficult to create a perfect result in a single generation, so adjustments using Image to Image (i2i) are necessary.
When generating with i2i, it’s recommended to adjust the mask area and pre-paint the parts you want to modify in the original image with the ideal colors using external software.
I got everything to work. My workflow is now on the level of gods. I fixed everything and it works with the latest version of ComfyUI. I even got it to work for an arbitrary amount of characters. In a singular day, I, with your black magic forbidden wizardry of a technique that only 100 out of a quarter million people have progressed in, feel I have made something precious like diamonds. I'm tempted to share, but my brain tells me I desperately need an ego check and that what I have is not even close to good enough to share. Over 136 version upgrades of the workflow within 4 months of progress off of 2 years prior experience and a life background in computer science is not enough for my brain to feel confident in sharing apparently. I'll give it another month or two unless someone actually reads this and messages me before then telling me to share.
Share it lol, I'll give it a try. Never hurts to put more workflows out there
Share. We want to be cursed by your black magic forbidden wizardry.
If you've managed to fix it to work with the latest version of ComfyUI, please share it with me. This is definitely something I need in my life.
https://www.youtube.com/watch?v=zhJJcegZ0MQ
I found this up sharing his workflow on youtube, doing the similar thing. It works with the new version Comfy. I tried with SDXL and my own character LoRAs, works great.
Can't for the life of me figure out inpainting, lol...I wish I wasn't too dumb to use this, if anyone ever makes a video explaining how to get 3 characters to work with this, shoot me a comment/message!
help me plz with this error:
La ejecución del prompt falló
Prompt outputs failed validation: LoadImageMask: - Value not in list: channel: 'clipspace/clipspace-paint-1180264.png [input]' not in ['alpha', 'red', 'green', 'blue']
:'c
Please verify that the channel of the Load Image (as Mask) node is set to alpha.
@kkr3 Do both masks have to be in alpha?
@Dante3rd
This error is likely occurring because a value other than the default has been entered for some reason.
It will not function properly if the value is anything other than [‘alpha’, ‘red’, ‘green’, ‘blue’]. This is a node specification.
@kkr3 Yes, I already changed both masks to alpha and it worked!
Thanks for your help and work!
If I have any more questions, I'll come back and ask you. n.n'