Open bit-9 opened 1 year ago
I dug a little into the code earlier. It doesn't actually read images as starting points for the pose manipulations. It has a set of preset coordinates to use as a starting point and it creates an image from them. The image box refers to the intermediate image that is generated directly into the "inputs" folder.
Basically the reason why is that it's much easier to create an image from points than to create points from an image.
In the short term, if you want to modify an existing OpenPose output from ComfyUI, I posted a temporary workaround modified from instructions from the creators of the comfyui_controlnet_aux repo in a related issue here: https://github.com/space-nuko/ComfyUI-OpenPose-Editor/issues/5
After selecting "Convert image widget to input" in the right-click-menu:
The provided node cannot be used:
I anticipate this feature could be useful for more robust workflows.