Mating-Press Sex
WAN2.2
Fresh off the RePress. Same set, ReCaptioned. ReTrained (in WAN2.2), render these sexy people, but only from close. Real Close.
mating-press sex,
A beautiful woman is {lying back|reclined} on a {white|black|pink|yellow|orange} {bed in a bedroom|couch in a living room} with her legs folded. A {dark-skinned||} man is pumping his {huge|large|small|} penis inside her vagina.
She has a fit body and a shaved vagina, her face barely visible from this angle,
{medium|} close-up view of the penetration action.Re: alpha (Jan 2026)
I originally trained this concept for WAN and HunyuanVideo. The HunyuanVideo version did not meet my standards, so I did not release it. One of the reasons I ended up charmed by WAN is how it seems to avoid mutations with much more consistency than HunyuanVideo, and, well, that's how the two LORAs differ. The HunyuanVideo one is a mess 3/4 times, but the uploaded LORA CAN hit the sweet spot, so it's not useless, and a fellow member said he wanted it so here it is. I have no plans to improve this one, it's a very limited POV with crazy occlusion going on...,
I trained this a couple weeks ago originally and it was a major fail. I've since isolated a much more consistent template style that I will share soon. It helped me unlock stability in weird/specific interactions like this. I hope!
mating-press sex
Tightly-framed view of of the penetration action, a beautiful {Russian|French|Swedish|Swiss|African|Latina|Austrian|Dutch|English|American|Siberian|Chinese|Ukrainian|Thai|Czech|Japanese|Indian|Caucasian|Spanish|Portuguese} woman lying back naked on a {bed|couch}. A {dark-skinned |||}man is pumping his {large|thick|giant|huge|glistening} penis into her vagina.
{bright|soft|warm candle|natural} lightingDisclaimer
Don't forget to breathe!
Description
FAQ
Comments (16)
The title has a typo
Seems like you've moved on to Wan now, which is understandable. But there are still a few Hunyuan/FramePack users out there. I'd love a Hunyuan version of this one! I have a few starting images that would be perfect to animate with this! No worries though if you've totally moved on at this point, just thought I'd throw it out there!
To be blunt, I've accepted the limits on my time and am being selfish now, focusing on LORAs I personally want to play with. I previously tried to train this LORA with HunyuanVideo and got a mess of mutations, but I'll see if I can rerun one on this more mature set and see what happens. Alternatively I am happy to share the dataset if anyone wants to give it a shot :)
@az420 Yeah, above all, you gotta do you. You've already done a great service to the community with these LoRAs. If you have the time and desire to give it one last shot, it would be much appreciated. But if not, again, totally understandable!
actually you know what? I think I actually did train this for HunyuanVideo... I just never got around to properly testing the checkpoints I extracted. With that said, I vaguely remember a latter epoch hitting the mark half the time. I'll try to dig that up!
@az420 Sweet! Even if it's not perfect, it's better than nothing. Most of the other LoRAs I have (not yours, haha) are pretty hit and miss, but it's fun to mess with them because you can still get some really good stuff when they happen to hit!
@Insult_to_Ninjary https://civitai.com/models/2203757?modelVersionId=2642555 so sorry about the delay, it requires a really good seed
@az420 I've been waiting for you to move on to Wan 2.2 :(
@honryindian he did, there is already a Wan version of this first, just change the model at the top! :)
@az420 no worries, you didn't even have to do it at all, so i really appreciate you digging it up! will give it a few test runs soon!
@honryindian i never say never :P
So far so good, at least for i2v. It took longer to get a non-cursed starting image from my Illustrious model than it did to get the video that I posted!
@Insult_to_Ninjary good to know, at least my crappy LORA is in good company then :P
Models tends to want to do faces properly, so this angle, with little/no/obscured faces, is easy to understand as a tough concept. Anyway, patience shines through!
@az420 don't be too hard on this LoRA, it at least seems pretty reliable for i2v, and once I learn how to prompt t2v better, I think it could be pretty decent for that as well, given some good RNG seed luck. I've tried 2 t2v gens so far, and they were both almost there, except for some wonky anatomy. I think the LoRA was definitely worth the share in the end!