Tried to create an pov kiss type of lora, with like only 20 pics but trained with 1900 steps.
As Dr. Ian Malcolm Said: Yeah, yeah, but you were so preoccupied with whether or not they could create a lora that you didn't stop to think if you should.
And since I have it why not share it, I guess.
Example Tags : Buchukisu, 1girl, solo, looking at viewer, open mouth, green eyes, purple hair, teeth, tongue, lips, saliva, makeup, drooling, close-up
2 Activation tags: Buchukisu and/or oral invitation
I used various styles but somehow the dominant style became from 阿黎 Pov kiss on patreon and pixiv, if you want to see more of their work, dataset also had art from Fellatrix, Rampage and Zxcv. But they show up rarely during generation, couldn't figure it out during my testing.
File came about big, but first time I used CivitAi lora trainer, other than the big file size, I want to say CivitAi lora trainer is very good, way faster than me spending computing units with Google collab. This is stupid, don't give me too much shit lol
Description
FAQ
Comments (12)
Lol this is so goofy
absolutely love you for this one you madlad
No crap will be given! I'd love to know a bit about how you make on with civit - with as maaaany training images as you'd like
I am not sure what you meant, but Civitai lora trainer says the file limit is 1000, this was my first and test lora making with Civitai so I dont know if they count the caption files, so maybe 500 images and 500 captions, or 1000 image if you manually caption them yourself, which would be annoying. It does cost 500 buzz but its very straight forward and easy, you just have to chose what kind of lora you are making, and dump your images, but you lose out on cutomization you would get if you were using the lora trainers on collab. https://civitai.com/models/22530 is required reading imo, I learned from reading this.
If I missed what you asked, just ask again, I don't know if you tried before or not.
@tigeracer Thanks! I'm reading that link next. Nah I was just joking about the number of images as I have no idea what's normal. I'm loving this LoRa, though! Messing with it now!
love it, is there a sd1.5 version?
sd1.5 would be appreciated
no effect,there is only the corresponding face orientation, but no corresponding exaggerated lips.I tried almost every keyword
could be the rampage lora, try without and see if it works, I just did it in 1 afternoon, just to try out the Civitai lora trainer with 30 image so it was basically 0 effort work from my part.
@tigeracer Actually, I spent the whole afternoon generating pictures with Ramepage's LORA, and these are the ones I picked that were barely effective. Do you have time to update this LORA recently?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.