Text-To-Image with Outfit Transfer Reference Image
Like with anything related to Krea Edit this is very much experimental but it works surprisingly well so I decided to share it with the community.
Notes:
Requires https://github.com/lbouaraba/comfyui-krea2edit or https://github.com/ostris/ComfyUI-Krea2-Ostris-Edit (See notes in Workflow)
Requires input outfit reference images in specific format. See example dataset: https://huggingface.co/datasets/AliveAi/outfits
Use trigger "transfer the outfit"
Workflow:
Two workflow options are attached. Let me know which one you prefer!
t2i_outfit_reference.json requires https://github.com/lbouaraba/comfyui-krea2edit - better accuracy but much slower
Krea2_Ostris_Edit_outfit.json requires https://github.com/ostris/ComfyUI-Krea2-Ostris-Edit - less perfect outfit reference adherence but much faster
Known limitations / issues:
Only trained on female outfits (probably also works for men)
Can sometimes produce images with two people. Re-try with different seed and/or update prompt to force "single person"
Description
FAQ
Comments (9)
How do you think is it possible to make it work for outfit swap?
Thanks for your work.
Are you using the 'two reference' mode of the identity edit workflow AND the identity edit LoRA in addition to this LoRA? Please provide some information so that we can try to align our output to something.
The exact workflow is attached
Please share your workflow
There are 2 WFs you can find in the "optional files" under the download area.
Is it really so difficult to read the author's notes carefully?
@strollvision I will contact you if I have any difficulties
its awesome lora, thanks a lot!
the model only can do t2i not i2i? I mean usea a image as base an change their outfit?