Trained at native 1024px without downscaling on a selection of "semi-leaning-towards-photo-realistic" high quality watermark-free Pony Diffusion V6-generated images, with resolutions of 1024x1024, 1024x768, and 768x1024 all represented on the dataset.
Recommended strength is around 0.6 - 0.8, but feel free to experiment with going higher or lower as you see fit.
In my testing, this Lora works quite well with both anime and realistic checkpoints due to the generally neutral style of the dataset. The checkpoint does NOT necessarily need to have any prior NSFW knowledge whatsoever.
Unordered list of relevant blowjob-related tags to use in any combination or amount you see fit:
`on knees, submissive female, 1girl, side view, face fuck, throat fuck, throat fucking, throat pounding, oral penetration, deepthroat, blowjob, fellatio, couple focus, penetration, female penetrated, dominant male, 1boy`
Unordered list of relevant doggystyle-related tags to use in any combination or amount you see fit:
`submissive female, 1girl, side view, doggystyle position, sex from behind, couple focus, penetration, female penetrated, dominant male, 1boy`
All showcase images were generated directly at the resolution they're displayed at, without upscaling.
Description
Initial upload.
FAQ
Comments (1)
Note that while I have enabled this for the online generator, this is an actual 1024px Lora and will likely give you better results generating locally directly at higher resolutions than CivitAI's generator allows (see the showcase for some examples).