About:
This LoRA has the purpose to (try to) replicate a in POV angle.
Be careful, as it sometimes generates body horror and disembodied .
It was trained in 768x768, using captions and 157 high-resolution images.
Btw, I also have no idea why the come out so big, in the dataset they were big, but normal. 😂
The demo images aren't edited, but they are cherry-picked and generated using hires-fix (0.3 ~ 0.4 denoising).
Suggested Settings for Inference:
Model: icbinpICantBelieveIts_newYear (but works with 99% of models)
Positive: cute young little woman, pov, , brunette, , , open mouth, half-closed eyes, airbnb room, short hair, sidelocks, dripping sweaty, body sweat <lora:povBlowjob_sd15_by_IsnAI:1:0.7>
Negative: verybadimagenegative_v1.3
Sampler: DPM++ 2M Karras
Steps: 28
CFG: 5~6
Resolution: 768x768, 512x768 or 768x512
LoRA Weights: 1:0.7
Note¹: As you can see, the LoRA tag has two weights: <lora:povBlowjob_sd15_by_IsnAI:1:0.7)
This is a feature of the A1111 (which many people aren't aware of), but you can set separate weights for the UNet and the Text Encoder. The TE is always a pain in the to get the training right, so it's usually better to reduce its weight in the inferences rather than having it undertrained.
Note²: You can get great image variations if you tweak resolution a bit.
Triggers:
Main: woman, , open mouth, man, (or black )
Tags Used on Training: , pov, teeth, half-closed eyes, closed eyes, wide eyes, legs, indoors, outdoors, pubic hair, , glasses, fingers, hands, , feet, hand on head, , makeup, , sunglasses
