Vore Body POV

Vore Body POVThese LoRAs consist in a first person view angle for vore generation. It has been trained on a mostly synthetic dataset for a semi-realistic style.The trigger keyword is: "POVVore"There are two files: POVVExpanded and POVVDeep; The first one has been trained to work better with anime styles and the second one is based on synthetic data that's more realistic in style.A good starting prompt structure for using this LoRA is (for YiffyMix v52, in this case): "uploaded on e621, [character], duo, human, vore, POVVore, vore, pov, first person view, prey pov, mouth closed, close up, ((almost fully inside)), [place]"And a good negative prompt is: "internal, organs, macro, micro, huge size difference, distorted, body horror, blurry, poorly drawn face, tentacles, deformed, taur, chakat, low res, tongue, teeth, open mouth, mouth open, foot, feet"Oddly enough, adding those last bits to the negative prompt tends to make image generation more consistent and prevent some artifacts.The examples for generation were ran with these paramenters:Model: YiffyMix v52Steps: 12CFG: 6Sampler: EulerScheduler: sgm_uniformClip skip: 2NOTE: The model tends to follow a somewhat realistic style. To try and achieve a more "anime" style I have used a mix of a few loras. To name:For a more realistic style: Just use the lora POVVExpanded with a weight of 0.8-1.15, you can play around with it changing strenghts or adding more loras to the mix as well.For a more anime style: A good mix that I've found is:POVVExpanded: 0.9POVVDeep: 0.35Pippuri style: 0.35Vore concepts: 0.5