CivArchive
    VegaNec Editorial V1 | Identity-Safe Krea 2 Realism - v1.0
    NSFW
    Preview 144970419
    Preview 144970418
    Preview 144970421
    Preview 144970425
    Preview 144970426
    Preview 144970427
    Preview 144970428
    Preview 144970429
    Preview 144970430
    Preview 144970431
    Preview 144970432
    Preview 144970434
    Preview 144970436
    Preview 144970438
    Preview 144970437
    Preview 144970439
    Preview 144970440
    Preview 144970441
    Preview 144970442
    Preview 144970443
    VegaNec Editorial V1 is a stackable editorial-realism LoRA for Krea 2.
    
    It is designed to add refined photographic polish while preserving the face,
    age, skin tone, body class, hair, and signature traits supplied by a prompt or
    a character LoRA.
    
    The model focuses on:
    
    - realistic skin microtexture without heavy plastic smoothing;
    - coherent editorial lighting and tonal separation;
    - natural fabric weight and cleaner material edges;
    - polished portrait, fashion, full-body, and mixed-light presentation;
    - compatibility with diverse adult character identities;
    - restrained strength so the character remains more important than the style.
    
    ## Trigger and strength
    
    **Trigger:** `veganec_editorial_v1`
    
    **Recommended strength:** `0.35`
    
    **Useful range:** `0.25-0.40`
    
    Start at `0.35`. Higher strengths are not recommended for general character
    stacking. In calibration, `0.55` and `0.75` began altering vulnerable hair
    traits, including Esi Mensah's ultra-short buzz cut and Thalenna Rook's cropped
    white curls.
    
    ## Character-LoRA stacking
    
    Recommended order:
    
    1. Load Krea 2.
    2. Apply the character LoRA at its normal strength.
    3. Apply `veganec_editorial_v1` at `0.35`.
    4. Prompt the character's identity and signature traits explicitly.
    
    Example:
    
    ```text
    character_trigger, exact character identity and signature traits,
    veganec_editorial_v1, photorealistic editorial photograph,
    realistic skin microtexture, coherent photographic lighting,
    natural fabric weight, balanced tonal separation
    ```
    
    For unusual hair or signature traits, state them directly. For example:
    
    ```text
    even ultra-short natural black buzz cut over the entire scalp,
    visible scalp contour, no ponytail, no bun, no braids, no wig, no extensions
    ```
    
    For hand-object scenes, use exactly one primary object and state hand ownership:
    
    ```text
    holding exactly one amber perfume bottle in her right hand,
    left hand empty and visible
    ```
    
    ## Validated compatibility
    
    Owner-reviewed benchmark coverage:
    
    | Character | Lane | Result at 0.35 |
    |---|---|---|
    | Nadia Vale | young modern influencer | PASS |
    | Hana Mori | mature Japanese artisan | PASS |
    | Esi Mensah | Ghanaian luxury professional | PASS |
    | Mirelle Vane | mature dark fantasy | PASS |
    | Thalenna Rook | young elven huntress | PASS |
    | Valeria Noctis | gothic vampire patron | PASS with one rejected hand-artifact frame |
    
    The benchmark used four fixed scene classes per character:
    
    - close identity portrait;
    - full-body editorial;
    - exactly-one-object hand/contact scene;
    - difficult mixed-light scene.
    
    Each scene was rendered character-only and character-plus-editorial with the
    same prompt and sampler seed. The owner passed 47 of 48 total outputs. One
    stacked Valeria object scene was rejected because the gripping hand had seven
    fingers; that frame must not be uploaded.
    
    ## Honest limits
    
    - This is not a character LoRA and does not guarantee recurring identity alone.
    - It is not a pose or composition-control model.
    - It does not guarantee perfect hands or zero anatomical defects.
    - Compatibility is proven only for the six tested character LoRAs and scene
      classes listed above.
    - Higher strength can overwrite hair or other vulnerable signature traits.
    - Use `0.35` as the tested default and preserve explicit character prompting.

    Description

    Training used 48 owner-approved, metadata-clean, identity-diverse editorial
    images. Benchmark characters were held out from training. The final benchmark
    tested modern, mature, fantasy, light/deep skin, slender/full/powerful/wiry body
    classes, close portraits, full bodies, hand-object contact, and mixed lighting.

    Details

    Downloads
    37
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/7/2026
    Updated
    10/7/2026
    Deleted
    -
    Trigger Words:
    veganec_editorial_v1

    Files

    veganec-editorial-v1-epoch-10.safetensors