One of recent loras got nuked by civit, you can download here on ko-fi, for free: https://ko-fi.com/luisacaotica
v2-------------------------------------------------------------------------------------------------
Better vagina anatomy, hell yeah, btw sometimes can do failure, but you can always use anothers loras like https://civarchive.com/models/1434650?modelVersionId=2012120 to increase fidelity.
follow exact the prompt : ZOOMPUSSY, the camera zooms out, she her laying down show her legs, and her outie vagina, naked, pussy, she keeps her hands down, don't touching anything, she is calm.
v1 -------------------------------------------------------------------------------------------------
Don't do high expectation!
i've tried hard, retrained again and again so much time, but no luck, i did merges ( a lot ) and no sucess at all, wan 2.2 doen'st know how genitals work, so, if you got some nsfw checkpoint, maybe you get luck.
so i will share the dataset (free porn!) for trainers [ already in 16 fps 5 seconds ] , so if someone can do better, i'll aprecciate that. [if you're trying to merge, remember, this model is rank 4 ]
Description
the high noise is merged with mystic and dr3aml4y
FAQ
Comments (5)
when someone releases a better version, it will be archived.
Wowee it zooms out to some thick thighs.
Been following your work for ages and I must admit, I like the spark of crazy genius you seem to have. You go girl!
@luisa_pinguin Well, I've made no less than 70-80 attempts, and among the various combos, I've managed to get good results using your lora in combination with this one: NSFW/Female Genitals helper for Wan T2V/I2V - v1.0 i2v | Wan Video LoRA | Civitai
Obviously, it doesn't matter, since half the work is done by the second lora, but this tells us a couple of things:
1. I changed the prompt to "ZOOMPUSSY, the camera zooms out and shows her belly, her legs, and her outie vagina" since outie vagina is the trigger for the second lora, but your lora still works 100% of the time and doesn't alter the subjects' faces, so that's great.
2. The second lora manages to get good results using the term "vagina," so at this point it might be worth trying it in the next attempt (if you're up for it).
For now, I have to stop testing, but tomorrow I'll bring you more updates because I want to help you with this, as I'm particularly interested in this lora.
you need a few closeup high quality in your dataset and make sure it is well captioned to say it is closeup and it should get detailed