Inpaint workflow V2
GUIDE FOR V2 WORKFLOW IS LOCATED HERE
06/12/2025
-Added ability to control mask opacity.
-Changed sampler to prevent problems with generation.
-Adjusted guide.
04/12/2025
-Added ability to do colormatch. Changing colormatch setting to 1 will try to match the color of the image.This is good if using VAE that changes color between encode/decode cycle or using different lora/model.
Older:
-Added differential diffusion
-Cleaned the workflow
-Created guide which can be found HERE
-This should work with 1.5 too, did not test.
Inpaint workflow XL V1.1
Due to request updated to work with XL. It is somewhat barebones compared to 1.5. Does not have segmentation or controlnet models currently.
I have a bit outdated comfyui, let me know if it is throwing some errors.
Description
This version uses efficient loader and can use loras with the inpainting. Totally forgot about that.
FAQ
Comments (14)
using 2.02_ALT but still got error:
Error occurred when executing Combine and Paste: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
what I'm doing wrong or what file I'd edit to fix this issue?
Hmm, interesting. Did you restart comfy and reload page after you installed everything?
@PotatCat restarted Comfyui, restarted pc, deleted python and copy back from original package of portable Comfyui (but only 3.11, not downgrading yet), SD1.5 with such error, SDXL not working at all with Ksampler efficient, other workflows works fine
@souljadogg Sadly i do not know anything about this specific issue, hard to start pinpointing what could be causing it :/
miss jinjarender scheduletomodel promttoschedule scheduletocond
Try alt version, it is using different system for prompting
potato
i know you have the furry maker embed
you cant hide it for long
i will hunt you down
give me the precious
Seems to be a bug for me. If I force the seed to be the same, it doesn't matter how much I change the ControlNet strength or base multiplier (even going from 1.0 to 0), it just spits out the exact same inpaint result like I never altered ControlNet. But if I change the seed, generate, then go back to the same seed from the prior image and try again, it now generates a different result like it should.
Edit- alt version does the same thing for me too. At least I noticed just changing the steps count by one seems to force it to regenerate properly, so I don't have to flip through seeds all the time.
Thanks btw, ignoring the annoying bug this is probably the best inpaint workflow I've found.
Yeah, this is from the efficiency sampler. If nothing changes from last generate it "freezes" the sampler and keeps the results. On earlier version there was option to choose the state of the sampler to hold/sample.
Sadly the controlnet settings do not trigger the "something has changed". I am annoyed by this also.
Thanks for the comment. Please leave review if you like, gives me energy to continue :)
@PotatCat I see, guess there's no easy workaround for that then, especially since it says the dev stopped maintaining the efficiency nodes. At least I know what the culprit is now, thanks.
@kosmo Ah, that is sad to hear. One of my favorite nodes ;(
Seems like comfy update broke efficiency nodes. As the original creator of efficiency nodes is not updating anymore i personally switched to jags111 fork https://github.com/jags111/efficiency-nodes-comfyui
The workflow makes its job great. But seriously, don't put the nodes that close next time. The pin too, I was wondering why I couldn't move the nodes to try and understand the flow of things.
The point of this whole GUI is to be able to see the whole process to modify it to each user case.
Thank you for comment.
It comes to personal preference how you want to have the nodes. For better user experience for new/non experienced users i decided to put them close and pin them. The idea was to make it as simple as possible and usable for even new users who have no experience with comfy. (～￣▽￣)～