CivArchive
    Lum from Urusei Yatsura (2022) - v1.0
    NSFW
    Preview 335534
    Preview 335533
    Preview 335530
    Preview 335529
    Preview 335525
    Preview 335524
    Preview 335523
    Preview 463070

    Lum from the anime Urusei Yatsura (2022). Trained on screenshots and conveys it's vibrant style, but obviously doesn't match up well with realism. However, you can try it with a weight of 0.7 and lower.

    The site already has Lum, but I wanted the anime style for myself.

    My first LoRa, so don't expect a miracle. I tried to use different poses and clothes to achieve a good result. Personally, I'm happy with what I got, because Lum looks great and conveys the style of the anime.

    Trained on Anything 3.0, Clip Skip 2.

    P.S. Negative promt affects the colors. Here's the negative promt that I use::

    (worst quality, low quality:1.4), EasyNegative, bad-artist, low res, (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation

    Description

    FAQ

    Comments (7)

    KangzMar 26, 2023
    CivitAI

    Good Shit friend, could you do Sakura from UY too?

    Kenfox
    Author
    Mar 26, 2023

    Glad you liked it! I hadn't thought that far ahead, but anything is possible :) For now, I want to see if I can improve Lora and expand the dataset of pictures to make Lora more flexible.

    KangzMar 26, 2023

    @Kenfox Ok, good luck friendo

    RomangeloMar 28, 2023· 3 reactions
    CivitAI

    Great one. Please make Ran, Oyuki and Mendou Ryouko, too.

    settima_aiApr 19, 2023
    CivitAI

    Fantastic LORA. Face is always so precise... Please, do also some male charater, like Ataru, Mendo, Rey... I'm planning to do a fanfiction about!

    1871594Jun 15, 2023
    CivitAI

    how did you get these to be so precise. mine never turn out so good

    Kenfox
    Author
    Jun 18, 2023· 2 reactions

    Hi! I did lora with kohya_ss, maybe that's it. I noticed that using Google Colab usually gives worse results.

    I used screenshots from the anime and some official art, selecting the clearest and displaying different emotions/clothes.

    In Automatic1111 you can see all the settings of the lora. I used one epoch and 100 steps, but as far as I know this is a controversial method and people prefer multiple epochs for better control. However, using multiple epochs usually doesn't give me the results I want.

    I didn't bother much with the settings, it was my first lora. Maybe the whole secret is just a good dataset.

    LORA
    SD 1.5

    Details

    Downloads
    4,170
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/26/2023
    Updated
    4/30/2026
    Deleted
    -
    Trigger Words:
    lum
    lum_bikini

    Files

    Lum.safetensors

    Mirrors

    CivitAI (1 mirrors)
    Other Platforms (TensorArt, SeaArt, etc.) (1 mirrors)

    Available On (1 platform)

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