A simple workflow that generates an image with Flux Dev (GGUF) and swap face with selected image, and refine the face using Flux.
Can also be used as an alternative to help make consistent characters.
Description
Added a mask helper to make face swap work better with hand, hair or anything else covering face.
FAQ
Comments (18)
It says AV_Facedetailer Node not found. Everything else seems to work... Any Idea why?
I'm on the same boat, have no idea
I spent 40 minutes and found the missing node. It is comfyui-art-venture by sipherxyz. You will find it on GitHub.
And then tons of mistakes. Missing this, missing that. F this s**t.
Use comfyui manager, install missing custom nodes, did that work?
@Kiyushiax no it doesnt find the note
@romanfirsov123437 Thank you
i tried replacing it with FaceDetailer of ComfyUI-Impact-Pack, it works
@romanfirsov123437 I also encountered the same problem. There was a prompt that the AV_Facedetailer node was missing. I couldn't find the answer on Google. Only when I came back here did I find that I had encountered it!
If u're not a dev. like me, install Comfyui with Pinokio.
@romanfirsov123437 u r the GOAT! Searched whole web but couldn't find it myself. I needed it, because the settings from the author of this model are much better, than if you just replace it with the stock values from Impact-Pack.
@romanfirsov123437 saved my day. Thanks
The error is not resolved. ㅠ_ㅠ
@aimodelfree i will update soon! im sorry
Update just released! Now everything will work properly and can be installed with ComfyUI Manager.
Thx for these perfect and clear workflows.
Easy and not 200 nodes... I like a lot.
I've just downloaded the Q8.gguff. Not sure that my very old 2009 HP Z600 + RTX 3060 survive but I'll try. :) For moment 250 s by image with your workflow is ok for me.
Working just fine, a bit slow but Ok. Thanks for the workflow my friend.
When I tried this workflow, pytorch 2.6 version raised unsafe pickle module.You must downgrade pytorch 2.5 or edit "SomePlace\Lib\site-packages\ultralytics\nn\tasks.py" belows
torch.serialization.add_safe_globals([DetectionModel]) ckpt = torch.load(file, map_location="cpu",weights_only=False)and the other thing is...
AV_FaceDetailer changed into comfyui-impact-pack FaceDetailer.
Update released! Just tested on latest nodes and all working.