Real Mutiple Reference Try On Workflow
Description
Add mask area
FAQ
Comments (20)
this is just the standard workflow with 2 tweaks. Your images show a masked image, but the workflow has no masks.
Yes, it is modified from standard workflow. But most people use standard workflow not with this small modifications
And mask is not important. It could do the same job using normal unmasked image
@xiaozhijason how to use it? It does only give me the picture of the clothes back.
@multimega download v1.2 and use mask editor on human ref. It would convert the mask to greyarea and the process would try to use cloth ref to fill in the greyarea
@xiaozhijason ok thanks. Im an new with these things. I masked the picture with a photoeditor, but it sounds like there should be a mask editor in the flow? I do not see that?
@multimega Mask editor built in in comfyui. Right click and scroll to find mask editor
Kontext is so insanely censored, it even tries to determine if you are trying to apply clothing to the input image that is 'too sexy', and then refuses! Its proper workflow method is to begin with a fully nude adult image, so requests to 'dress' are reducing the nudity- a conceptual direction Kontext is hard trained to approve of. The opposite direction, and Kontext tries to prevent the change.
great insights. will try it out a bit later.
uncensored versions are here already
@TurboCoomer where
Thank you, I just started messing with Kontext and thought I was going in sane when "Place a naked man in the middle of the room." kept giving me an old guy in a 3 piece suit. Guh, I find it completely offensive that they're shipping these models out with a built in robo-nanny to make sure we don't get too racy with our creations!
TurboCoomer Oh, thanks mate, didn't see your reply.
this workflow is dowgrading a lot depencies. destroyed new build
Might be you could view the workflow in RH and build the own version of yours. I was in the latest comfyui updates and work fine.
in general just work with multiple separate portable builds or else you're in version mismatch hell
是有效的
why it dont change
pls help me , it not error
https://ibb.co/fd7RTfDK
The latest version of Comfy has a different GetImage node, you may want to update the workflow.
Ran it 3 times with the human ref's clothes masked out, it didn't replace it with the supplied garmet.
It resizes the human ref's image, this will likely lead to image degradation. I'd suggest a new workflow that maintains the human ref's image size exactly.