CivArchive
    Preview 705671
    Preview 705674
    Preview 705673

    Lulu from Final Fantasy X!

    Now on ANIMA, NoobAI-XL (Epsilon AND V-PRED), PonyXL and SD1.5!

    ANIMA tags:

    Lulu \(ff10\), 1girl,

    red eye, hair over one eye, one eye covered,
    hair ornament, hair stick, black hair, single hair bun, multiple braids, purple bead necklace, red necklace, pearl necklace, mole under mouth, eyeshadow, purple lips,
    curvy, large breasts, cleavage

    Lulu's outfit:

    off-shoulder dress, black dress, fur trim, multiple belts, black thighhighs, thigh boots

    optional: belt skirt, holding stuffed toy, holding moogle, holding sabotender

    Style control: 3d, painting \(medium\) - to swap between 3d and 2d styles.

    SETTINGS FOR ANIMA:

    My main guide for Anima gen: https://civarchive.com/articles/30357/about-anima-settings

    Also my preview can help!

    NoobAI-XL / PonyXL tags:

    Main appearance:

    Lulu \(ff10\), 1girl, mature female,

    red eye, hair over one eye, one eye covered,
    hair ornament, hair stick, black hair, single hair bun, multiple braids, jewelry, purple bead necklace, pearl necklace, mole under mouth, eyeshadow, purple lips,
    curvy, huge breasts, cleavage

    Lulu's outfit:

    bare shoulders, off-shoulder black dress, fur trim, multiple belts, black thighhighs, lace thighhighs,
    For NoobAI-XL: off shoulder dress, black dress

    Optional:

    you can remove "red eye, hair over one eye, one eye covered" and write "red eyes" if you want to see 2 eyes without hair bang.

    NoobAI-XL quality settings:

    For NoobAI-XL V-pred:

    Check my guide about V-pred image generating!

    The best model to use V-pred is Cyberfix! But pure NoobAI-XL V-pred is also good.

    For Illustrious mixes and NoobAI-XL Epsilon: use NoobAI-Xl Epsilon (especially mixes based on NoobAI-XL Epsilon!).

    Guide about NoobAI-XL Epsilon.

    PonyDiffusion v6 (XL) quality settings (updated):

    Style settings (very important):

    score_9, score_8_up, score_7_up, score_6_up, score_5_up, source_anime

    You can remain only "score_9" if you use style lora because all scores except score_9 supress style loras.

    Negative prompt (recommended):

    3D, lowres, censored, mosaic censoring, bar censor, blurry, bloom, bad anatomy, deformed, head out of frame, watermark, worst quality, low quality, jpeg artifacts

    Version 3 SD 1.5 animefull:

    Features:

    • 220+ dataset;

    • Smaller filesize.

    • Tags were reworked manually in Dataset tag editor for easy prompting. So read new tokens for better experience.

    Main info:

    Trained on animefull;

    0.8-1 is the best radius.

    Clip skip 2.

    Tags SD1.5:

    Main appearance:

    Luluffx, mature female, 1girl,

    black hair, single hair bun, hair over one eye, one eye covered, lipstick, makeup, purple necklace;

    optional: braid;

    Default dress:

    off-shoulder dress, black dress, strapless, fur trim, multiple belts, garter straps, thighhighs

    Recommended:

    curvy, large breasts, cleavage,

    perfect eyes, perfect face, detailed face, sweat

    Stable Diffusion 1.5 settings

    Prompt for VERSION 3:

    Main appearance:

    Ingrid, 1girl, mature female, dark skinned female, pink hair, long hair, hair intakes, lipstick, purple lips, makeup, yellow eyes, mole under mouth,

    Description

    SD1.5.ckpt based version for realistic checkpoint models (Stable Diffusion1.5, deliberate, dreamshaper, etc).

    Clip skip 1.

    FAQ

    Comments (6)

    jenxiMay 6, 2023· 1 reaction
    CivitAI

    The realistic version seems to give the same results as the anime version.

    Enigmata
    Author
    May 6, 2023

    Realistic version is more for clip skip 1 models but yes it also can do anime / cartoon-like arts.

    AraraMay 7, 2023· 1 reaction
    CivitAI

    Heya was just wondering, from the preview and trying them out, are the uploaded files for realistic and anime loras reversed? as in the wrong way round?

    Enigmata
    Author
    May 7, 2023· 1 reaction

    It's funny but no it is not reversed :3

    I trained alt. version on vanilla SD1.5.ckpt with clip skip 1 without any NAI-based models and generated little preview on deliberate and dreamshaper models. Since these are universal checkpoint models, as you can see, they can generate both anime-like illustrations and realism.

    So I decided to make a version for NAI-based checkpoint models and universal ones.

    kalitbartmanJul 5, 2023· 3 reactions
    CivitAI

    can't get v1.1 to look like the first screenshot even when using generation data

    Enigmata
    Author
    Jul 17, 2023

    I suppose it depends on version of SD and xformers. I generated it without xformers. Now after torch 2.0 update SD works very slow without xformers and everybody use it. But with xformers SD can't repeat picture which was generated without xformers.

    LORA
    SD 1.5

    Details

    Downloads
    1,174
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/6/2023
    Updated
    11/2/2026
    Deleted
    -
    Trigger Words:
    LuluFFX

    Files

    LuluFFX_SD15.safetensors

    Mirrors

    HuggingFace (1 mirrors)
    TensorFiles (1 mirrors)