I've just finished the initial release of my nodes that I've been working on.
Currently only one, but I think it's pretty special. I've been using it a lot with z-image-turbo w/ control net and Qwen-Image-Edit-2509 but it works great with anything that takes a pose obviously.
Interactive Pose Editor
Category: ae-in-workflow
An interactive pose editor that allows users to manipulate OpenPose keypoints with full control.
Features:
Individual Keypoint Editing: Click and drag any keypoint
Person Selection: Automatically selects the person closest to your mouse
Multi-person Support: Handle scenes with multiple people
Pose Manipulation: Move, rotate, and scale entire poses
Smart Spacing: Automatically spaces overlapping people
Keypoint Fixing: Automatically estimates missing keypoints using T-pose proportions
Duplication: Clone poses for creating variations
Caching: Remembers your edits for faster subsequent sessions
Controls:
Left Click: Select and drag keypoints
Left Drag: Move keypoint
Ctrl + Left Drag: Move keypoint with children
Middle Click + Drag: Move selected person
Scroll: Zoom selected person
Ctrl + Scroll: Rotate selected person
Shift + Scroll: Rotate children around nearest keypoint
Ctrl + D: Duplicate selected person
Ctrl + X: Delete selected person
Ctrl + N: Add new T-pose person
Ctrl + F: Fix missing keypoints
Ctrl + R: Flip horizontally (mirror pose)
Ctrl + Shift + R: Flip horizontally (turn around)
Ctrl + Z: Undo last action
Ctrl + Shift + Z / Ctrl + Y: Redo action
Ctrl + O: Reset to original input
ESC: Save & Exit
Inputs:
pose_data (optional): OpenPose format pose data as JSON string (if not provided, starts with default T-pose)
padding (optional): Padding around poses in output (default: 128)
seed (optional): Change to force re-execution with same pose data (default: 0)
reset_cached_window_position (optional): Reset window position/size to defaults (default: False)
Outputs:
image: Rendered pose visualization image
edited_pose_data: Modified pose data in OpenPose format as JSON string
Description
FAQ
Comments (4)
how does it work as i can't get it to work at all
Usual steps to install:
Clone the repo to your custom_nodes folder.
pip install -r requirements.txt with the ComfyUI venv activated.
How to use:
(Optional) You can use Control Net preprocessors to extract a "starting pose"
You can also just have the node on it's own to start with a default t-pose.
Read the controls, it supports some pretty intuitive pose editing.
## Issues
- Window doesn't show - Try toggling the reset window position cache.
- I'm getting errors or the node won't load - Check your terminal for errors related to missing requirements or whatever else and let me know. If you are certain you have installed it correctly and it's not working, let me know your setup (standalone or source, windows/linux/mac, etc.)
- The pose isn't working in my ZImage/Qwen/Wan/etc. - Sorry but this is unrelated to this tool, just a workflow problem.
@by_ae did that , wasn't straight forward but i got it to work , i just assumed i only needed to download the workflow and comfy will do the rest , not opening github , and the ComfyUI venv inside the ae-in-workflow folder or i can copy the requirments text folder to the main directory as i am a bit rusty on all that stuff
Interesting, it works for me, I just need to experiment more. This opens up several possibilities. Thank you.