CivArchive
    Slop Bounce WAN 2.2 I2V - WAN 2.2 I2V High
    NSFW

    Edit 9-12-2025: Added the LoRAs as separate I2V High and Low for people using Runpod or using the online generator or something.

    Baked a slop LoRA as a test a week ago on my 3090 while waiting for Bouncing Boobs to be updated to WAN 2.2 I2V. Trained sloppily on 7 of my old AI generated videos that were using WAN 2.1 Bounce LoRA using PGC's modified GUI of Kvento's Musubi Tuner GUI with a minor fix for WAN 2.2 I2V training that just made the --i2v toggle only apply to latent cache creation instead of training. Trained for 20 epochs but the uploaded epoch is number 5.

    Videos were lazily manually tagged like
    A white haired woman wearing a micro bikini bounces causing her breasts to bounce, shake, jiggle and sway.

    So prompting breasts bounce or jiggle will generally work. Not prompting anything specific also sort of works. Prompting something like

    her breasts bounce, shake, jiggle and sway

    will work.

    Training data includes the modified musubi_tuner_gui.py for PGC's GUI with the fixed toggle for properly caching latents for i2v as well as the .json and .toml files that contain the dataset settings and training settings. Also has the workflow images for the model page's videos although the videos should also have the Comfyui workflow included for once.

    Description

    Split download for people using Runpod or the online generator?

    FAQ

    Comments (20)

    Natsu24Sep 13, 2025· 3 reactions
    CivitAI

    Can you make TI2V-5B version please? 🫶

    _Kapot_Sep 13, 2025· 5 reactions
    CivitAI

    Thanks for dividing the fires in low and high. It's quite helpful for people using Lora Manager

    necosanderSep 19, 2025· 4 reactions
    CivitAI

    First of all, thank you for all your help.

    I trained on the dataset you shared

    on 16GB of VRAM (4070tis).

    High and low were trained separately.

    The high was 2 hours and 19 minutes.

    Low was 2 hours and 17 minutes.

    I had to use t2v because an error occurred, but I was very satisfied with the results.

    (I just realized that I should have replaced py.)

    anonimoose
    Author
    Sep 19, 2025· 2 reactions

    I'm glad the training data was useful to someone.

    YinsuTarvoSep 29, 2025· 2 reactions
    CivitAI

    amazing you are the best!! Is there a possibility that you will make one for twerking?

    anonimoose
    Author
    Oct 2, 2025· 2 reactions

    I just uploaded Slop Twerk.

    A_Large_CranberryOct 2, 2025· 4 reactions
    CivitAI

    This Lora can also be used for twerking, not always perfect but I like it. Starting with an image from behind with stuff like "she raises and lowers her hips repeatedly" along with your prompt for jiggles and bounces

    anonimoose
    Author
    Oct 2, 2025· 1 reaction

    I saw your work. Very creative use of the Slop Bounce LoRA to get a twerking motion. I just uploaded Slop Twerk so hopefully that works a bit better.

    A_Large_CranberryOct 3, 2025· 1 reaction

    @anonimoose good to know! Always nice to see more jiggles, thanks for sharing your loras!

    InoSimOct 3, 2025· 1 reaction
    CivitAI

    I wanted to reallyyyyy thank you man !
    You made the only workflow that actually keep the identity of the original picture.

    Big big thanks !!!

    OriginalSinXOOct 10, 2025· 1 reaction
    CivitAI

    Amazing Lora, a pleasure to work with continues to surprise me! ❤️‍🔥

    Simon_BelmontOct 18, 2025· 4 reactions
    CivitAI

    An honestly underrated and underused lora around here. This really helps with "jellyfication" and bouncey movements, even in other situations, coupled with other loras, put it at the end of the lora chain, and it will simply "liven" your flesh lumps.

    Highly recommended!

    bakaboy1234Oct 23, 2025· 1 reaction
    CivitAI

    what model did you use to make the ref images that are animated in your samples? is it IL?

    anonimoose
    Author
    Oct 31, 2025· 1 reaction

    Mostly with iamddtla's Pornmaster-Pro noob-V3.
    https://civitai.com/models/1045588?modelVersionId=1767015

    The Hishi Amazon and dark elves were Pornmaster-Pro noob-V5. The Vocaloids, Hilda from Pokemon, and Takarada Rikka were using a custom merge between YNSrealv3 and Pornmaster-Pro noob-V3.

    furuteerikaOct 31, 2025· 1 reaction
    CivitAI

    I love the image style you made and i am struggling on reproducing it. Would you mind posting one of your ref images so that i could check its workflow? thx a lot!

    catsimboythepoop123Nov 2, 2025· 1 reaction
    CivitAI

    This is a great "seasoning" lora which adds a more jiggly jiggle to other bouncy loras.

    UltraFreshNov 29, 2025· 4 reactions
    CivitAI

    Shaddup and Take My ...Support

    zirtapodDec 7, 2025· 1 reaction
    CivitAI

    Does this work with T2V?

    sb129Jan 19, 2026
    CivitAI

    one of the best and most versatile wan loras ive tried yet! thank you so much! love the style and jiggle

    KitsuneVixenJan 26, 2026· 2 reactions
    CivitAI

    This just changed everything for me. Thank you!

    LORA
    Wan Video 2.2 I2V-A14B

    Details

    Downloads
    13,186
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/12/2025
    Updated
    5/14/2026
    Deleted
    -

    Files

    bounce_test_HighNoise-000005.safetensors

    Mirrors

    HuggingFace (45 mirrors)

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.