CivArchive
    Blowbang | Double blowjob WAN 2.1 I2V - K3NK - Wan i2v v1.0
    NSFW

    Someone requested a bounty with only 7 days, you have my version here..

    Double Blowjob Gangbang LoRa for Wan i2v

    • Trained at 258 on a dataset of 81clips 1sec to 5sec (several source videos), 5e-05 Learning Rate, Num_Repeat = 2

    The workflow im using is:

    * you can find a slightly modded wan workflow inside the sample video's metadata

    Example prompt:

    d0ubl3bl0wj0b, A woman is kneeling on the floor between two men. There are two penises directly in front of her. The woman is performing a deepthroat blowjob on the penis on the left/right while grabing the penis on the right/left while moving her hand back and forth.

    ***** This LoRA is shared without warranties and with the condition that it is used in a lawful and responsible way. I do not support or take responsibility for illegal, harmful, or harassing uses. By downloading or using it, you accept that you are solely responsible for how it is used. *****

    Description

    FAQ

    Comments (10)

    aniyaqinqinJun 27, 2025
    CivitAI

    I want to ask, is your learning rate much higher than 2.4, I also want to train LoRa, but I have a lot of datasets, and the bigwigs recommend 2E4 and I don't know how to set it

    K3NK
    Author
    Jun 27, 2025

    It depends on if you need high detail, also lower rate means slower training

    CharlieBrown0115Jun 28, 2025

    QUICK QUESTION< DOES THE GUYS HAVE TO BE ALREADY ON THE IMAGE? OR THEY ARE GENERATED BY THE LORA? AND HOW YOU KEEP FACIAL consistency


    ilya808rolf994Jun 27, 2025
    CivitAI

    Sorry for being off topic and writing through a translator. Does anyone know a hint describing hand actions: stroking, touching, caressing, etc. No matter how hard I tried to achieve this, the results on the video are such that I want to laugh out loud. Finding a stallion or rhinoceros penis on this site is not a problem, but finding truly useful things that can create a quality erotic game is impossible.

    K3NK
    Author
    Jun 27, 2025

    Have you tried with my workflow? Or other workflows, try using kijai's workflow too

    ilya808rolf994Jun 27, 2025

    @K3NK Unfortunately, I don't work with the "vermicelli" program, but only with WanGP

    rustybutlerJun 27, 2025

    Use a spline editor and vace motion control. You'll be able to have control of the hands that way.

    ilya808rolf994Jun 27, 2025

    @rustybutler Did you write this tip for those who work with wan in stable diffusion ComfyUI?

    Tezozomoctli2Jul 1, 2025
    CivitAI

    Do you accept commissions?

    I have attempted to train a Wan lora using musubi-tuner but I don't know what I am doing wrong so I'm thinking of just giving up and just ask experts to do a commission instead.

    So I tested created a lora with 2-3seconds long videos at 500x480p and at16fps. I only trained using 3 videos. I know that is nearly not enough videos to train on, I just wanted to get a sense of how long the training process would take. And the training process took about 24hrs.

    That can't be normal right? 24hrs with only 3 videos to train on? Do I need to reduce the pixel sizes of my videos even more? I know I didn't share my config.toml file with all the details so it is hard to guess, but I am really just stumped trying to find out with my training would be that long. Also, I am using an RTX 5070 Ti Super 16vram.

    So that's is why I'm thinking of just going around asking others if they do commissions lol.

    K3NK
    Author
    Jul 1, 2025· 1 reaction

    try diffusion-pipe and use lower res with longer bucket sizes

    LORA
    Wan Video 14B i2v 480p
    by K3NK
    Download (Beta)
    style

    Details

    Downloads
    5,001
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    6/27/2025
    Updated
    4/21/2026
    Deleted
    4/16/2026
    Trigger Words:
    d0ubl3bl0wj0b