CivArchive
    BJD Doll Style - v1.0
    Preview 1

    Description

    Doll Style / BJD Photography LoRA

    [CN] 这是一个基于 Illustrious 架构训练的 LoRA 模型,专注于生成 BJD (Ball-jointed doll) 娃娃摄影风格的图像。它能带来独特的树脂质感和精致的娃娃妆容。

    [EN] This is a LoRA model trained on the Illustrious architecture, focused on generating BJD (Ball-jointed doll) photography style images. It delivers unique resin textures and exquisite doll makeup.


    如何使用 / How to use

    v0.2:

    • Trigger Word: dollstyle
    • Recommended Weight: 1.0
    • CFG: 3.0

    ⚠️ 关键提示 / Key Notes


    1. 🚫 关于质量提示词 / About Quality Tags

    • [CN] 请勿使用正面质量提示词(例如 masterpiece, best quality 等)。
    • [EN] Do NOT use positive quality tags (e.g., masterpiece, best quality).
    • [Reason]: 质量提示词本质上带有强烈的特定画风权重,会破坏本模型带来的 BJD 摄影质感。
    • [Reason]: Quality tags essentially act as style modifiers and will interfere with the BJD photography texture intended by this model.

    2. 🎨 底模兼容性 / Checkpoint Compatibility

    • [CN] 本模型与大多数 Illustrious 类底模兼容性良好。如果你使用的不是写实/真实系底模(或者想要更强的“非人”娃娃感),建议使用 1.0 以上的权重,并在负面提示词中加入 anime。
    • [EN] Compatible with most Illustrious-based checkpoints. If you are using a non-photorealistic base model (or want a stronger artificial doll look), sugges* ***ng a weight above 1.0 and adding anime to the Negative Prompts.


    3. ⚙️ 参数设置 / Settings (CFG, Steps)

    • [CN] CFG 和 Steps 等参数可能需要根据你选择的底模进行微调。具体设置请参考示例图片的元数据(Metadata),图像包含 ComfyUI 工作流信息。
    • [EN] Parameters like CFG and Steps may need tweaking depending on the specific checkpoint. Please refer to the Metadata of the example images for specific settings; they contain ComfyUI workflow information.

    4. ✨ 高清修复 / Hires Fix

    • [CN] 强烈建议开启 Hires Fix,这对于还原 BJD 的皮肤纹理和眼部细节非常重要。
    • [EN] Highly recommend using Hires Fix to achieve the best resin skin texture and eye details.

    Details

    Downloads
    4
    Platform
    ShakkerAI
    Platform Status
    Available
    Created
    6/17/2026
    Updated
    6/17/2026
    Deleted
    -
    Trigger Words:
    dollstyle

    Files

    dollstyle.safetensors

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)
    ShakkerAI (1 mirrors)