CivArchive
    RoxyF - V1
    NSFW
    Preview 108329250
    Preview 108329251

    # grok_girl – Hyper-Realistic Blonde Character LoRA

    ## What does your model do?

    grok_girl is a character-specific LoRA trained on 6 high-resolution, photorealistic images of a single blonde woman with a sharp bob haircut, athletic build, and seductive/playful expressions.

    When triggered with *grok_girl**, it injects her exact likeness (face, body proportions, skin tone, hair style, and signature vibe) into any scene while preserving the base model’s style.

    ---

    ## What is it for?

    - Perfect 1:1 character replication in new poses, outfits, lighting, or environments

    - NSFW & fashion photoshoot generation (lingerie, latex, cosplay, editorial)

    - Video animation via Hunyuan/Wan (smooth motion from static poses)

    - Cinematic storytelling with a consistent female lead

    - Stacking with style/motion LoRAs (e.g. + cinematic lighting, + dynamic camera)

    ---

    ## What is your model good at?

    | Strength | Details |

    |--------|---------|

    | Face consistency | 99% identity lock across angles, lighting, expressions |

    | Body accuracy | Exact bust/waist/hip ratio, toned legs, posture |

    | Hair & makeup | Sharp blonde bob, freckles, lip shape, eyeliner |

    | Outfit adaptation | Adapts corsets, latex, bikinis, lace with correct fit & texture |

    | Pose flexibility | Sitting, crawling, lying, standing — all retain character |

    | Video-ready | Trained on Wan/Hunyuan — natural motion interpolation |

    ---

    ## What should it be used for?

    - Commercial NSFW/fashion content (with proper licensing)

    - AI filmmaking (consistent actress in every shot)

    - Character design prototyping (test outfits, settings, moods)

    - LoRA stacking workflows (combine with realisticVision, cinematic, latex, etc.)

    ---

    ## What is your model bad at?

    | Weakness | Why |

    |--------|-----|

    | Diverse ethnicities / ages | Trained on one Caucasian woman ~25yo — will override other faces |

    | Male characters | No training data — results in feminized males |

    | Complex hands / fingers | Small dataset → occasional deformation |

    | Extreme angles (top-down, worm’s eye) | Limited

    Description

    Text 1:1 einfügen

    LORA
    Wan Video 14B t2v

    Details

    Downloads
    78
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/30/2025
    Updated
    4/28/2026
    Deleted
    -

    Files

    RoxyF-000007.safetensors

    Mirrors

    CivitAI (1 mirrors)

    2361859_training_data.zip

    Mirrors

    CivitAI (1 mirrors)