trigger word : ohwx woman
recommended checkpoint : sdxl 1.0 base
recommended lora strength : 1.0 for simple prompts, 1.11/1.22 for complex prompts
my first real person/celeb SDXL lora,
after spending much money on cloud gpus and after many hours of training/ failed attempts, I think I have got something worth posting to civitai (or is it?, pls let me know in the comments, positive/negative just any kind of feedback is appreciated <3)
I didn't bother testing it on any stylized merge as I am satisfied with the base model results, which can be great, when correctly prompted.
if you like my stuff and wish I research and train more SDXL loras pls consider supporting me on Kofi
or on buy me coffee
Description
FAQ
Comments (7)
Very nice results! I'm sorry I was about to post my version. At least it's from a different year.
yess, post it, always welcome to see how better models can get,
I have mostly used prompts from your felicity model hehe,
love your sdxl models <3
@dogu_cat Thanks, I'm a big fan of your 1.5 models. You going to SDXL great news for the community.
@dogu_cat @razzz have you guys though about creating a lora for younger Emma Watson, from around 2001-2002, Philosophers Stone and Chamber of Secrets era? I'd be so awesome <3
@Norby123 It will be misused, and it's also against Civit ToS
each photo has a different "someone else" mixed in. Is this SDXL messing with the likeness? 1.5 celebrities look way better than SDXL
its a new thing, me and some others are still experimenting with it,
SDXL loras will only get better with time, like any other new stuff