CivArchive

    Wan 2.2 update. The preview images were all created using the lightning loras. The workflow can be downloaded in the "training images" download.

    The main part of the prompt is:

    A woman is straddling a man and having sex with him. She bounces up and down on top of the man

    Depending on the angle of the camera, you can either directly see the genitals or not. You can add this to your prompt to try to get them to show up:

    You can see the woman bouncing up and down on the man's erect penis.

    You can prompt for different angles:

    {The view is from the front and at a 45 degree angle|The view is from the side|The view is from behind at a 45 degree angle|The view is POV}

    By default, the woman's knees will probably be pointed down, so if you want more of a squatting position with her knees up, you can prompt for that this way:

    She is squatting with her knees raised

    Description

    FAQ

    Comments (24)

    gurilagardnrMar 15, 2025· 5 reactions
    CivitAI

    I struggled the first couple runs, but the i2v trick is to turn down the lora strength. 0.7 was the sweet-spot for me.

    Kung_fu_PronMar 15, 2025· 4 reactions
    CivitAI

    anyone found hard to make it work with i2v for a man out of frame, or if the girl is on her side with one of her legs being held up, trying to figure this out.

    NoneofYourBusiness2Mar 15, 2025· 1 reaction
    CivitAI

    Love it, thanks. If there's material for hip movements other than up and down, that would be highly appreciated (as I think such way for woman to fuck is even hotter)

    dtwr434
    Author
    Mar 15, 2025

    The training material had some back and forth stuff in it, captioned with "She grinds back and forth on top of the man", so you can try that, but I don't think it actually listens. I think loras can only really learn one type of movement at a time as far as I can tell.

    @dtwr434 Awesome, I will try, and if I learn something or get it to do this, I will report back with the info.

    1172769Mar 16, 2025· 3 reactions
    CivitAI

    lots of progress then the previous version, but this version seems that women are mostly busty

    fronyaxMar 21, 2025· 1 reaction
    CivitAI

    any plan to make for 1.3B one??

    dtwr434
    Author
    Mar 21, 2025· 2 reactions

    It seems unlikely at the moment. I think the 1.3B model seemed to know way less about NSFW stuff, so I'd imagine it would not turn out very well. Maybe if I get my 14B loras in a state I'm happy with, I'll try a 1.3B version just to see what happens.

    fronyaxMar 21, 2025· 1 reaction

    @dtwr434 the "better titfuk" lora has 1.3B one and the nsfw is pretty good in my opinion.

    astral_divaMar 22, 2025· 7 reactions
    CivitAI

    For some (good) reason, this cowgirl LoRA makes almost any other position in I2V work better. It adds a general bounce/jiggle motion that has improved videos of other positions, not just cowgirl. It also seems to make the model understand nsfw movements better. I haven't looked at the dataset you used, but this is quickly becoming my "all-purpose" nsfw Wan LoRA. Good job! 💙

    Theeditor1Apr 14, 2025

    Hey so what workflow are you using with I2V since this persons is T2V? Sorry im new and havent worked a lot of stuff out yet. thanks

    astral_divaApr 15, 2025· 1 reaction

    @Theeditor1 Hey, I am using a customized version of this workflow: https://civitai.com/models/1278171?modelVersionId=1468785 Beware though that you should use the default Comfy Core LoRA loader and not the one that comes with this workflow, i.e., don't use the "Power LoRA Loader (rgthree)" node, since when I used it it messed up a lot of my generations by making the frames turn gray and faded. I2V works great for me with this one otherwise.

    TartrixroarMar 23, 2025· 3 reactions
    CivitAI

    I am using this lora with Wan 14B and it basically does not work. Woman does not move her butt up and down. I tried adjusting a bit of the strength. I doing it with i2v (cowgirl images) and it just doesn't seem to work. The woman is moving but being still. Help?

    dtwr434
    Author
    Mar 23, 2025

    I think there are a lot of tools and workflows floating around that don't seem to work well with loras, or at least some loras. The best I can suggest is to download the workflow from the training images download provided here and use that as a starting point for your workflow.

    TartrixroarMar 23, 2025

    @dtwr434 I am only finding one for t2v one, not for i2v, any help?

    dtwr434
    Author
    Mar 23, 2025· 1 reaction

    @Tartrixroar Oh yeah, for I2V, I haven't really tested it or done much with it. I used this workflow when I briefly played around with it, and it seemed to work for some of my loras. I would assume it should work for all of them. https://civitai.com/models/1297230/wan-video-i2v-upscaling-and-frame-interpolation.

    _crz_Mar 29, 2025
    CivitAI

    What scheduler sampler and cfg do you use? Your generations look good and every workflow I download uses something different.
    Or do you have a particular workflow that you use?

    dtwr434
    Author
    Mar 29, 2025· 2 reactions

    The "Training Images" download is actually just my json workflow, if you want to try it.

    _crz_Mar 29, 2025

    @dtwr434  Wouldnt have thought to check that zip. Thanks

    TartrixroarApr 2, 2025· 6 reactions
    CivitAI

    I am still having trouble with this lora. The woman barely moves. It works but not consistent enough to be usable. Considering it takes over 10 minutes per generation and I get shit results, this has become tiring to use. Do you by any chance have a i2v workflow? I would really appreciate it. Not a t2v, an i2v as I don't know how to link nodes yet. thanks!

    dtwr434
    Author
    Apr 2, 2025· 1 reaction

    I use a modified version of this workflow with some success: https://civitai.com/models/1297230/wan-video-i2v-upscaling-and-frame-interpolation?modelVersionId=1517031. However, I do feel like models trained specifically for I2V just work better for I2V. I imagine I will eventually get around to training a proper I2V version at some point, but it's possible it's just not worth it until then. I trained this mostly with T2V in mind.

    makiaeveliApr 14, 2025

    LOL it takes that long for everyone. You're right to ask for workflows. But it's more or less been confirmed that I2V trained models and T2V models don't work amazingly interchangeably. It seems it's completely dependent on the training setup. There are several other I2V sex models that work well

    SD_AI_2025Apr 14, 2025

    "Do you by any chance have a i2v workflow?" You have hundreds if i2v workflows embedded in most of the videos posted on Civitai. Drag and drop in comfyUI.

    Or use Pinokio. 1 click install of Wan. Simple to use interface (without nodes). You can do t2v, i2v, image A to image B, etc. Works great (that's how I use Wan).

    negative1ne483May 9, 2025

    i tried it with an i2V, and it worked perfectly. excellent results and good generations. takes time for the first generation, but all of them do. repeated ones are much quicker.