CivArchive
    BVB - Big Veiny Breasts - v1.0
    NSFW
    Preview 119266605
    Preview 118916823
    Preview 118651559
    Preview 118651551
    Preview 118651558
    Preview 118651552
    Preview 118651548
    Preview 118651554
    Preview 118651555

    BVB (Big Veiny Breasts) | v1.1.1 Update

    Version 1.1.1 focuses on enhanced realism and dynamic range. This isn't just a "one-size-fits-all" texture anymore; you now have granular control over the intensity, pattern, and physical scale of the effect.

    What’s New in v1.1.1

    • Improved Realism: Better skin integration and depth—less "painted on," more "under the skin."

    • Size Versatility: Now supports a range from small breasts to huge, maintaining anatomical accuracy.

    • Pattern Control: Introduce specific visual styles through prompting.


    Prompting & Patterns

    You can now define the specific "look" of the vascularity. Try mixing and matching these keywords:

    • Size Range: Use small breasts, medium breasts, or huge breasts to scale the bust size.

    • The "Faint" to "Intense" Look: Prompt for faint for a subtle, naturalistic realism. Prompt for prominent for high-visibility, high-detail results.

    • The Pattern: Prompt for spider-like vein pattern or branching vein pattern for different patterns.


    Usage Tips

    To get the most out of this LoRA without "deep-frying" your entire image, follow these guidelines:

    • The "Detailer" Method (Recommended): For the cleanest results, apply the LoRA during an Adetailer or Inpainting step focused specifically on the chest. This keeps your character's face and background untouched by the LoRA's texture.

    • Recommended Weights:

      • 0.7: Best for subtle, natural skin textures and "barely there" realism.

      • 1.0: Maximum intensity. Expect very prominent (often blue/vivid) vascularity.

    • Trigger Word: [bvb]

    • Samplers: * Euler / Euler A: Best for smooth, photographic skin.

      • DPM++ 2M: Best for maximum "crunch" and hyper-detailed skin texture.


    Compatibility

    • Base Model: Optimized for z-image, but provides good results on Jib Mix v1 and other realism-focused checkpoints.

    • Goal: Adds realistic vein patterns, enhances skin depth, and supports "science experiment" or hyper-realism aesthetics.


    Creator’s Note

    Thanks for the support on the first version! v1.1.1 is a direct result of fine-tuning the feedback I received. I’m constantly looking to improve the realism of these textures, so please post your generations in the gallery! Seeing your results helps me calibrate the next version.

    Description

    First version, based on close-up shots of the chest, which may affect your base generations. 

    FAQ

    Comments (9)

    pushpiblFeb 3, 2026· 1 reaction
    CivitAI

    或许你应该涂抹掉数据集的脸部,来减少对面部的干扰

    pushpiblFeb 3, 2026· 1 reaction

    人脸风格被干扰的像ins网红欧美女性了

    austinlele
    Author
    Feb 3, 2026

    "Perhaps you should smooth out the faces in the dataset to reduce interference with the facial features."

    Seems like a good suggestion, can you provide examples of the distortion that you are seeing? What does your workflow look like? Thanks

    I will say the dataset contains very minimal full faces, but perhaps the few that exist are over taking.

    pushpiblFeb 3, 2026· 1 reaction

    @austinlele 我发了一张对比图,从左到右依次是:1.无任何Lora,2.只有bvb,3.只有人像Lora,4.人像Lora+bvb

    austinlele
    Author
    Feb 12, 2026

    @pushpibl i dont see the images posted, but i am working on a new version without faces at all

    Soup_FusionFeb 3, 2026· 1 reaction
    CivitAI

    peak. I've attempted this so many times and have failed every time lmao. how are you getting the veins to train properly? they're such a fine detail that it's never learned it correctly. also are you planning on training for the base model as well?

    Soup_FusionFeb 3, 2026· 1 reaction

    also would you be able to do more bulging veins? like where the vascular lines are raised a bit from the skin.

    austinlele
    Author
    Feb 3, 2026· 1 reaction

    Probably luck and some cherry picking haha, but ive done a lot of modifying of the dataset manually, and set the LR higher than normal, as you can see by file size. that seemed to help a lot.

    I can / would train on base, but could use pointers as i havent had good results yet..

    austinlele
    Author
    Feb 3, 2026· 1 reaction

    @Soup_Fusion I can try my best ha, will take me some time tho

    LORA
    ZImageTurbo

    Details

    Downloads
    298
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/25/2026
    Updated
    6/11/2026
    Deleted
    -
    Trigger Words:
    bvb

    Files

    bvb_v1_3.safetensors

    Mirrors

    TensorFiles (1 mirrors)