So, I was training models locally, and the newest AI-toolkit update broke Krea2 training for me :(
I figured I would try out the on site Lora training since I have the better part of a year of blue buzz saved up on a meme concept I was working on. A post by @Experimental_dAIver really struck me by how true it was. this one: Image posted by Experimental_dAIver
I'm sure a lot of people noticed that its many of the same people on the Featured Image board each day, so I thought I'd make a Lora trained on the specific type of images that get featured, because for a while it really was just exactly what was laid out in that image lol.
This was really a bare minimum effort shitpost kind of Lora, but it actually worked out kinda good. I did naural language captions prefixed with the trigger word Featured_Fishin and MAYBE a user tag if they showed up alot in the featured images.
I didn't spend enough effort to actually try to capture any user's specific style, but you may see hints of them if you tag for them. maybe I'll make a better effort version with a more curated list of images.
All example images used on strength 1.0, which I reccomend you use.
Here are the users I specifically tagged for:
@Cinnadust @eileen33bits @Experimental_dAIver @Geragl @Lady_Luminous @LumaRift @purplelady @Vladaa
They were captioned like this: "user_cinnadust", "user_eleen33bits" (yes, I know I fucked up their username on accident in the caption. oopsie), "user_ExperimentaldAIver", "user_Geragl", "user_Lady_Luminous", "user_LumaRift", "user_purplelady", "user_Vladaa".
Maybe with this Lora I may achieve my lifelong goal of reaching the civitai front page, and maybe you can too. (I should probably learn what the golden ratio is first)

