This LoRA contextually fills in the green area in image1 with head from image2.
Basic prompt:
fill in the green area in image1 with head from image2
Sometimes it helps to expand the basic prompt a bit:
fill in the green area in image1 with head from image2, maintain correct anatomy (maintain identity, from back)
Mix with other face swap or consistence LoRA sometimes helps sometimes not. The green area should be specifically RGB(0,255,00). You guys can do it with Mask Editor in ComfyUI (latest), but personally I convert mask to area with RGB(0,255,0).
Best resolution for image2 (source head) is 1024 x 1024. There are no rules about background, zoom nor pose of the source head.
Workflow i'm using:
https://civarchive.com/models/2307457/aint-another-head-swap-workflow?modelVersionId=2596183
Description
FAQ
Comments (14)
Brilliant lora. Lots of stuff are possible if you get creative.
Thank You! There will be more of "green" LoRA.
@orzechowy3334318 And I'll keep checking them out. I mean, this one is perfect.
Could you show (or upload) the whole workflow, please? I workflow image you posted is, I think, only partial.
Thank you
I'm trying the GGUF model and it seems to be working up to adding the green mask to image 1 but the Process group ends up with a totally black image. Any ideas?... ** OK, I found the answer in the comments for your workflow - bypass SageAttention and it works. **
wow! how did you do this as a lora??
Determination ;)
It's pretty cool. But often the head size changes, and you can't keep the expression of the 1st image
The less strength used, the more original expression is maintained. But this is not a goal of the lora to keep expression. The goal is to inject the head into photo context.
@orzechowy3334318 Yeah if I understand it correctly the edit model + sampler only sees the green paint, not the original underneath. There's no way it can keep the original expression - but maybe one could extend the workflow to crop the original face and generate a pose outline with the face and input that with the edit model as a reference? That should also help with head size. And maybe adding a consistency lora could also help? Just thinking - haven't tried it.
@EricRollei21 You are free to modificate the workflow as you wish. I'm going to publish v2.2 soon and there will be some clues for your needs. My goal is to gain CONTEXTUAL head pose with expression. However you can extend the prompt for call desired expression. It works.
@orzechowy3334318 I can confirm that prompt describing the face will affect the end result.