>>66693062>>66578151LS anon here, it was inded with comfy
>>66676751kek
nta, came out as funny as I imagined
>>66670277great stuff
>>66671196yuri banzai
Is hug
(and further action) from behind possible with this?
I know the model itself likely can, but will regional prompt work is the question
>>66693694for SDXL, I'd say try training on base SDXL with OFT and make sure to set the constrain (try the highest from the paper: 1e-05)
base SDXL should have enough anime in it, it just needs to be strong armed a bit
also, make sure to set different learning rates for the the text encoder and UNet
I think UNet needs lower lr as it fries earlier (though OFT doesn't fry very easily)
