CivArchive
    pov cowgirl position (hunyuan video) - v1.0
    NSFW

    The censorship is added by myself after generation due to Japanese law.

    This lora is trained on realistic style videos.

    Sample prompt:

    high quality cinematic video of a cute girl having sex with a man. She moves her hips up and down intensely. She is moving intensely. She is straddling on the man. cowgirl position, pov, sex, penis, pussy, vaginal, uncensored, extremely cute girl with long hair, pale skin, realistic, photorealistic, cinematic lighting, cinematic video, ultra-highres, highres, 4k, absurdres, masterpiece.

    Description

    FAQ

    Comments (25)

    LatentDreamDec 21, 2024· 1 reaction
    CivitAI

    is this trained on images or videos?
    looks like is trained on video, not static images, wich is great in this case.
    What are the trigger words?

    yanagi099
    Author
    Dec 21, 2024· 3 reactions

    My training caption is "a video of a girl having sex with a boy. cowgirl position, pov". Due to the limitaion of frame number to train, it is hard to train long-term motion. I reccomend to add "She is moving intensely." or something into your prompt.

    sdimgDec 21, 2024· 1 reaction

    @yanagi099 Great stuff. What else are you planning to train? Would be good to see various positions like doggystyle from the side and with her looking back at camera.

    LatentDreamDec 21, 2024· 1 reaction

    @yanagi099 its ok. Few frames are enough for the model to learn movements. I would suggest to specify in description if the lora is trained on images or video sequence, so users knows exactly what kind of bias and behaviour to expect from the lora😏
    thanks for info and for sharing your work.

    yanagi099
    Author
    Dec 22, 2024· 1 reaction

    @LatentDream You are true. Thank you!

    Dobo_JDec 22, 2024· 1 reaction

    @yanagi099 Could you speak about where you did the training, for those of us who'd want to try our own.

    yanagi099
    Author
    Dec 22, 2024· 4 reactions

    @Dobo_J I trained using diffusion-pipe (https://github.com/tdrussell/diffusion-pipe.git) on wsl2. See the page for details. Using my RTX 4060 TI 16GB, I trained on about 8 videos in resolution 244x244 and 17 frames.

    Dobo_JDec 22, 2024· 1 reaction

    @yanagi099 Thanks Yanagi!

    TsterbtaDec 24, 2024

    @yanagi099 How long did it take to train for this lora?

    Dobo_JDec 22, 2024· 1 reaction
    CivitAI

    More details for meta data would be great. Size? CFG? Steps? Flow_shift? num_frames? fps? Maybe if you can share workflow if used with comfyui. Thanks for the share, will try it out!

    guy33Dec 22, 2024· 3 reactions
    CivitAI

    How much VRAM is needed to run this thing? I have RTX 4060 TI 16GB VRAM. Is it enough?

    yanagi099
    Author
    Dec 22, 2024· 2 reactions

    It is enough in some ways. I was able to train a lora with 16GB vram on about 512x512 resolution for images, and about 240x240 resolution and 17 frames for video. If you want, you can train on higher resolution and frames, but it will take so much time. Due to this limitation, it would be hard to train details on something and long-term motion on 16GM vram.

    To train this lora, using my RTX 4060 TI 16GB, I trained on 8 videos, on 240x240 resolution and 17 frames.

    guy33Dec 22, 2024· 1 reaction

    @yanagi099 Got it! Then I think LoRa training should be done on cloud services like runpod.io where you can rent multiple powerful GPUs like 4xH100. One H100 costs $2.50/hour

    DiodamyrDec 23, 2024· 15 reactions
    CivitAI

    This is absolutely killer, please keep making these, lol ^^

    KaizendDec 23, 2024
    CivitAI

    Great Lora and even works with anime. If you don't mind could you share what you tagged your training images/videos for this?

    DomDomTomTomDec 25, 2024· 6 reactions
    CivitAI

    This works surprisingly well for blowjobs as well due to the range of motion, just so you know lol

    TurboCoomerDec 31, 2024· 1 reaction

    for missionary too! it alters prompt description not so much even at 1.0 weight

    zczcgDec 31, 2024· 1 reaction
    CivitAI

    good lora!!!I have a small question,cuz the face is similiar,how can I change the face?can by connect other lora or FaceID or EcomID

    hadengJan 3, 2025
    CivitAI

    Workflow?

    cuyjinJan 7, 2025
    CivitAI

    Great, one question: is this model faster than CogVideo img a video? I have a 12GB 3060 so I don't know if I can use it.

    WillieLonkyaJan 29, 2025

    There are workflows/models for lower Vram cards

    OFIAJan 30, 2025
    CivitAI

    how did you train the lur, Con Diffusion-Pipe u otro?

    krisscaldicott530Feb 11, 2025· 2 reactions
    CivitAI

    please could i get a workflow for this?

    ClocksmithFeb 23, 2025· 2 reactions

    This will at least give you a prompt if you download the video and drag it to comfyui. I think the critical part is "first person pov, cowgirl position sex."
    https://civitai.com/images/59076429

    MortalounetAug 15, 2025
    CivitAI

    Hello, I am new in AI photo generation, and looking for someone who could teach me how to generate videos like the one above, and I pay you for this service of course ;) thank you!