There are a lot of I2V Blowjob loras but I hadn't seen one that starts with a penis by itself
trained on [[512x768],[768x512]]x97 frames @ 16fps
made with 89 POV videos that start with a solo penis, goal is to take a dick pic and watch someone suck it, this doesn’t undress people so it should be fine TOS wise
I used dick, cock and penis interchangeably
can be used to spawn one or two women
lightening lora seems to work fine, all examples should have workflow
Description
FAQ
Comments (23)
You should've trained the LoRA such that it understands a combined input image with the initial starting image on one side and the "reference image" on the other side - if you know what I mean...
I don’t think this would able to stay on the site
Use the FFGO lora with it, it's exactly what you're asking for.
@Ohmlaash What Lora is that?
@Ohmlaash hehe, I was pretty much the first one to successfully create a FFGO video on kijais github. But I'm not sure if it is the right tool for this. Thanks for reminding, will try it
@Rosettasees FFGO-Video-Customization on github.
@Ohmlaash Noice! 🗣️ New toy to play with "FFGO"... Thanks!!!
@premiumfp I thought so because you are limited to a 720p input image. Exactly half of it gets lost for the environment image. If you add two additional references, you end up with a quarter of the input image. You can't fit a lot in there that could help you with consistency.
@anonymist should be fine TOS wise actually, doesn't undress anyone. Otherwise any regular BJ I2V lora would get banned too
@Ohmlaash how do use that? just get the two loras and load them to the lora list, what about the iamges reference setup ..do you hav eworkflow I can look at?the
@premiumfp @irishpc491 A few months back I used Veteran AI's native workflow as a base and modified it to my own liking with a bit more advanced stitching and my usual optimizations. If you look up his video on YT I'm sure you can figure it out easily, the workflows are in the description.
@premiumfp @fenasikerim if you combine a background remover with crop by mask nodes with 0 reserves on all sides, you can get better results as the initial/background image can be around 60-70% of the whole canvas and you can stitch the other component according to their aspect ratio, like a horizontal car on the very top, a vertical standing person and the background next to each other under it, this way you maximize the size of the references and get negligible quality loss, that you can gain back with a detailer node and some sharpening and upscaling. Just be specific with the orientation of the references in the prompt.
@Ohmlaash FFGO seems to work. Gotta test it a little bit. Mind providing a POV starting image? xD
@fenasikerim Suuure! Here you go fam ;D https://litter.catbox.moe/tmdpo275dz065lcz.gif
Works well, nice and simple concept
Thanks for this. Amazing work with this one. Any chance you'll do one with the same concept but the woman doing cowgirl? That would be awesome.
Probably eventually, but it's a lot of effort where I can't really socialize with anyone, and I think all the gooning is starting to desensitize me. So it might be a while before I am up to it again
@anonymist No worries, I understand!
Works well combining with a character Lora (for the woman).
Great, now I can disappoint even virtual women with my small PP
why wouldnt' this make more sense to be t2v? Female character isn't even in frame.
because I have a dick I can take a pic of and no one around to suck it? I2V is literally the only way starting with just a dick makes sense
@anonymist If it was possible to make it a specific woman that came into frame, that would be awesome.