CivArchive
    Moria Luluka / 森亜るるか (Meitantei Precure! / 名探偵プリキュア!) - Illustrious - v1.0
    NSFW
    Preview 142795309
    Preview 142795320
    Preview 142795308
    Preview 142795317
    Preview 142795322
    Preview 142795312
    Preview 142795311
    Preview 142795332
    Preview 142795323
    Preview 142795315
    Preview 142795331
    Preview 142795316
    Preview 142795321
    Preview 142795324
    Preview 142795328
    Preview 142795327
    Preview 142795329
    Preview 142795330

    森亜るるか — Moria Luluka / 自前LoRA(2026-09-10学習)

    Meitantei Precure! (名探偵プリキュア!), 2026 TV anime (Toei Animation).

    ★ Just the trigger word is enough - her black dress, white collar and frilled apron

    come out without describing them.

    RECOMMENDED

    Weight : 0.9 (1.0 starts to change the dress / 0.8 looks the same as 0.9)

    Base model : Illustrious (WAI-Illustrious SDXL v1.7)

    Sampler : euler_a, CFG 5, 30 steps

    Size : 1024x1344, SINGLE PASS (no hires fix)

    PROMPT (core)

    luluka_moria, 1girl, solo

    FULL BODY - add these to keep her hairstyle

    (long bangs:1.25), hair between eyes, (wavy hair:1.2), very long hair

    ICE CREAM (her thing - it is in the training frames)

    (holding ice cream cone:1.3), (eating ice cream:1.2)

    CASUAL CLOTHES (outfit changes work)

    (cream knit sweater:1.2), pleated skirt, brown boots, casual clothes

    (white hoodie:1.2), denim skirt, sneakers, casual clothes

    NEGATIVE

    worst quality, low quality, blurry, bad anatomy, bad hands, extra digits,

    text, watermark, signature, 3d

    TRAINING

    82 images (x5 repeats) / 8 epochs / 1640 steps

    unet lr 0.0001 / text encoder lr 5e-05

    trained at dim 64, then resized to dim 32 (217MB) with kohya resize_lora -

    compared side by side with the dim 64 file, no visible difference.

    trained on waiIllustriousSDXL_v170.safetensors with kohya sd-scripts

    Screenshots from the TV series, machine-filtered (ccip) then checked by eye.

    Station logo (top 9%) and burned-in subtitles (bottom 12%) were cropped off.

    No AI-generated images in the set.

    TAGS THE LORA ACTUALLY LEARNED (from ss_tag_frequency)

    luluka_moria 82, parody 58, blush 52, looking at viewer 39, virtual youtuber 29,

    anime coloring 27, ribbon 27, sky 25, open mouth 24, closed mouth 23, portrait 23,

    cloud 21, holding 19, jewelry 19, day 18, food 16, dress 15, black dress 14,

    smile 13, outdoors 13, necklace 11, holding food 10

    WHAT THAT LIST TELLS YOU (the honest weak points)

    ★ No `full body` / `standing` tag at all. In full body shots her bangs and hair ends

    drift - that is why the hair tags above are there.

    ★ Back views tend to add a red ribbon on the dress that is not in the official design.

    ★ The purple mascot is in many training frames and sometimes appears on its own.

    ★ `parody 58` / `virtual youtuber 29` are WD14 misreading the art style.

    I checked all 82 frames by eye - there is no other character mixed in.

    ★ `food 16` / `holding food 10` are her ice cream scenes.

    ## ■ 説明(日本語)

    『名探偵プリキュア!』の森亜るるかです。推奨0.9。

    ★**トリガーだけで黒いワンピース・白い襟・フリルのエプロンが出ます。**

    ★全身の絵では前髪と毛先が変わりやすいので、髪のタグ(上の FULL BODY)を足してください。

    ★アイスを食べる場面が素材に入っているので、アイスはよく出ます。

    ★私服への着替えも効きます(上の CASUAL CLOTHES)。

    ★後ろ姿では公式に無い赤いリボンが出ることがあります。

    ★紫のマスコットが素材に多く写っていたので、勝手に出ることがあります。

    ★dim 64 で学習し、dim 32(217MB)に縮めています。縮める前と並べて見比べ、見た目の差はありませんでした。

    Description

    LORA
    Illustrious

    Details

    Downloads
    35
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/14/2026
    Updated
    9/16/2026
    Deleted
    -

    Files

    luluka_moria.safetensors

    Mirrors

    CivitAI (1 mirrors)