CivArchive
    Preview 29780770
    Preview 29780807
    Preview 29780820
    Preview 29780822
    Preview 29780829
    Preview 29780857
    Preview 29780871
    Preview 29780891
    Preview 29780894
    Preview 29780909
    Preview 29780929
    Preview 29780951
    Preview 29780958
    Preview 29780969
    Preview 29780984
    Preview 29780990

    Yoko Tsuno character, from the same name graphic novel by Roger Leloup.

    This Lora is intended to generate her only, I've had poor results trying to integrate the others.

    Lots of pre-processing with this one. Few usable images, and old posters to restore. But it's working - just you'll always get her 1980-1990 look. No other adaptation has been made at this point of writing and there are even fewer fanarts so it will stick to the graphic novel style no matter what.

    I honestly have no idea what could be the best model(s) for this one, but i guess any semi-anime will do. I've had pretty good results with PorcelainEarthMix so sticking with this one. It also seem to be the easiest to control face rendering with.

    • Trained Word: Yoko Tsuno

    • SFW only - integrating NSFW in the dataset may degrade overall quality, especially face

    • Dataset size: 44 images

    • Can draw both her Earth outfits as well as bodysuits/spacesuits, very good at adding space helmets without any other lora/lyco

    • Can change outfit color palette (to some extent)

    • Reinforcement tags: Short Black Hair, Hairband, Traditional Media, Science Fiction, Retro Artstyle, 1980s \(style\), 1990s \(style\)
      For XL, you can also use Graphic Novels Illustration, especially for Cheyenne for which it is a style trigger keyword.
      For Earth outfits: Red Dress, Black Pants, Kimono, Obi, Cheongsem (this one is unstable)
      For space outfits: bodysuit, spacesuit, space helmet, futuristic

    • Optimal weight range: 0.9 - 1 (1.5) / 0.7 (XL)

    Some of the samples try to draw distant relatives of the scenes from the original, the others are completely random mashups. Maybe you shouldn't play with Monya's time machine in secret...

    (Bows are hard, mkay ?)

    Extra resources used not listed in generation data:

    For SD 1.5:
    Unlimited Blade Works: https://civarchive.com/models/69323
    Bowser: https://civarchive.com/models/7300
    Princess Peach: https://civarchive.com/models/16250
    Tron Style embedding: https://civarchive.com/models/4317


    For XL:
    Envy Scifi Streamline: https://civarchive.com/models/233464/envy-scifi-streamline-xl-01
    Archeress (bow helper): https://civarchive.com/models/275606/archeress
    XL Lava Lands: https://civarchive.com/models/194401/xl-lava-lands
    Underwater Scooter: (unpublished, not sure i will, it's bad)

    Hope you like it, and have fun :)

    Description

    Tested with Flux.1 D. Lora has close to no effect at standard (1) weight on NF4.

    FAQ

    Comments (10)

    NanashiAnonSep 18, 2024
    CivitAI

    Can you do a Pony version?

    ze_thriller
    Author
    Sep 18, 2024

    Doable, but just so you know, i've never used Pony before.

    NanashiAnonSep 18, 2024· 1 reaction

    @ze_thriller As far as I understand it, Pony doesn't need anything particularly special to train and is just throwing the same training data as XL with different settings. Someone with more experience training other SD branches and Pony likely knows more than me though.

    ze_thriller
    Author
    Sep 18, 2024

    I trust you'll do better than me - which isn't going to be too hard lol.

    For now images are, well, very strange to say the least. The XL version will work there too but i can't say how good.

    NanashiAnonSep 18, 2024

    @ze_thriller I don't have the training data.

    ze_thriller
    Author
    Sep 19, 2024· 1 reaction

    @NanashiAnon I meant generation-wise. I don't put to doubt your training skills, but method/settings are the same as for XL version so we should have rather close results right ?

    For now prompt following is, well... close to non existent. I refuse to post a model with samples that would belong in a fail gallery.

    See here: https://civitai.com/posts/6774117

    If i manage somehow to do better then we're in business !

    Pony has always been out of scope because of its strangeness, even more now Flux is out.
    You requested it and i fell into it for reasons, now i'll try to deliver :)
    But this may be a one-time experience.

    ze_thriller
    Author
    Sep 24, 2024· 2 reactions

    @NanashiAnon it took me a bunch of trial and error but i'm starting to have something decent. Now to adapt the prompts and that should do it 😊

    NanashiAnonFeb 9, 2025

    @ze_thriller Had forgotten about this. Thanks. For whatever it's worth, Illustrious can use same training data with slightly different settings.

    ze_thriller
    Author
    Feb 10, 2025· 1 reaction

    @NanashiAnon Working on it actually. It's a little less of a headache than with pony.

    NanashiAnonOct 12, 2024
    CivitAI

    Some moderator seems to think 20 something university graduates with breasts are underage and hit a picture of her.

    Edit: Was restored. Had already deleted other pics with LoRA out of abundance of caution.

    LORA
    Flux.1 D

    Details

    Downloads
    210
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/16/2024
    Updated
    5/14/2026
    Deleted
    -
    Trigger Words:
    Yoko Tsuno

    Files

    yoko_flux_v5.safetensors

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)

    Available On (1 platform)

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