CivArchive
    Sheena & Mimi (Anima) - v5
    Preview 147928235
    Preview 147928376
    Preview 147928375
    Preview 147928383
    Preview 147928384
    Preview 147928395

    Character LoRA for Sheena and Mimi from "Kimi ga Shinu made Koi wo Shitai" (きみが死ぬまで恋をしたい), trained on the Anima architecture.

    Both characters are in a single weight file — one download, two characters.

    IMPORTANT: Base model is Anima (anima16 / 29BBETA / aniAnimeColoring). This LoRA does NOT work on SD1.5, SDXL, Pony, Illustrious or NoobAI.

    ## Trigger Words

    | Character | Trigger |

    |---|---|

    | Sheena | totsuki_sheena |

    | Mimi | kagari_mimi |

    ## Recommended Settings

    - Steps: 30

    - CFG: 3.0

    - Sampler: er_sde

    - Scheduler: beta57

    - LoRA weight: 0.75

    - Resolution: 1024x1536

    ## Prompting

    Anima uses Qwen3-0.6B as text encoder and reads natural language, not weighted syntax. (tag:1.2) does nothing — write full sentences.

    A working prompt needs four parts:

    1. Quality: best quality, masterpiece, score_7, safe

    2. Character: trigger word + outfit

    3. Composition: upper body, looking at viewer, smile

    4. Style: flat cel shading, clean thin black lineart, muted colors

    Without the style block you get a white-background ID photo.

    ## Character Notes

    Sheena — short light-brown bob, black X hair clip, green eyes, navy sailor dress with large white collar edged in two navy stripes, white front placket, long balloon sleeves, cinched waist. White knee socks, black round-toe mary janes.

    Mimi — very petite, long wavy pink hair past the waist, a single ahoge, violet-magenta eyes, loose navy sailor dress with the same white collar, oversized sleeves hanging straight down.

    Sheena is the taller one. If both are in frame, state who stands where — the model defaults to equal-height symmetry otherwise.

    ## Known Limits

    - Two-character physical contact: upper body is reliable, lower body falls apart. Use upper body / cowboy shot for duo poses.

    - Hands: heart-hands style poses need 2-4 seeds to land clean.

    - 2.9B base models (29BBETA / anima29B) load the LoRA fine but lose the collar detail — the white placket and two buttons turn into a large bow. Use 2B bases for accuracy.

    ## License

    Fan-made LoRA for personal, non-commercial use. Character designs belong to the original rights holders.

    Description

    LORA
    Anima

    Details

    Downloads
    19
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/4/2026
    Updated
    11/7/2026
    Deleted
    -
    Trigger Words:
    totsuki_sheena, kagari_mimi

    Files

    kimishinu_v5.safetensors

    Mirrors

    CivitAI (1 mirrors)