I wanted to work on my throated lora but was having problems getting the lora to understand clothing or being more flexible. This is V1 of me testing with a different data set using joycaption. V2 will have exactly the same prompts on all images,but this seems to work well enough for me to think about sharing it. It was just using whatever joycaption spat out.see image for prompts, it might handle other stuff I just don't want to sit testing every possible variation. It works with Acornisspinning as well however I had better results on that chkpt using the two weeker steps with lower weights (0.6 and 0.45 or something to that effect) also seems to work ok with character loras. go wild feel free to leave feedback in case I missed anything.The prompts look like this try chopping and changing everything in the brackets you can change high-angle for above, not sure if it it knows low angle ? you can change locations, i was just lazy making preview images.also you don't need a watermark I just thought It would be funny to see if it trained them in.{high-angle} photograph depicting a (woman) ,kneeling in front of a man,performing oral sex on a man,his legs and lower abdomen visible.with the man's large erect penis in her mouth,sucking on the large penis,indoors,modern british livingroom,the mans penis centre of frame.she is looking at the viewer with a lustful gaze,,long hair,(water mark in the bottom corner reads "<a target="_blank" rel="ugc" href="http://Badlora.com">Badlora.com</a>")

Yet another blowjob lora V1

blowtest v1_000004000.safetensors

4000 steps

2750 steps

3050 steps

https://civitai.com/models/960922/yet-another-blowjob-lora-v1?modelVersionId=1075884

Files with SHA256: add47ae5110633d0a53cef96b24ea5a97e5171e85785d0afc9befb6db81b9062