Produces pretty decent hands for some reason
Great synergy with my Fei and Masami Chie LoRAs.
Description
FAQ
Comments (4)
u can use SDXL as images and then train those images on 1.5, might help improve the quality of your 1.5 LORAs
dude what
For your preview images, the Model metadata shows they were created using inpainting... why isn't this txt2img? very sus, and misleading information as to what the model actually does.
My workflow is:
raw output -> hires fix -> multidiffusion upscale -> inpaint to sharpen faces/hands or w/e minor detail I find lacking.
The lora is loaded in every step of the process and the preview images I posted represent the style just fine.
I'd get your point if this was a concept lora or w/e, but this is a style lora, so it really doesn't matter.
edit: here's a raw output grid if you care https://imgur.com/a/HER9s37