CivArchive
    Moss-PaxSinica - v1.0
    NSFW
    Preview 148353281
    Preview 148353696
    Preview 148353842
    Preview 148353990
    Preview 148353999
    Preview 148354107

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    盛世兴替

    心口相传

    福泽万邦

    天下大同

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    Moss-PaxSinica是一款基于MOSS-Image SFT打造的融合模型。

    在 Dirty-Moss 基础上进行融合和微调。

    在此感谢作者以及前置贡献者的付出与努力。

    两个版本中

    Moss-PaxSinica_V1 细节丰富,肤色白皙,融合东西方审美于一身

    Moss-PaxSinica_V2 细节丰富,灵动飘逸,更加符合东亚审美

    建议参数:

    2160x2880分辨率以下任意组合

    强烈建议使用 ComfyUI-Spectrum-MOSS 自定义组件与 comfy kitchen attention进行工作流优化

    避免使用任何加速Lora

    采样器推荐采用 er_sde

    调度器推荐采用 simple

    生成步数 30~40步

    cfg 3~4

    建议配合以下两款TextEncoder使用:

    qwen3.5_9b_heretic_v2_int8_hq.safetensors

    qwen3.5_9b_ultra_uncensored_heretic_int8_convrot.safetensors

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    The Rise and Fall of Prosperous Eras

    Passed Down from Heart to Heart

    Bestowing Blessings Upon All Nations

    A World of Universal Harmony

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    Moss-PaxSinica is a merged model built upon MOSS-Image SFT.

    It is the result of merging and fine-tuning based on "Dirty-Moss".

    Special thanks to the original authors and all previous contributors for their hard work and dedication.

    Available in two versions:

    Moss-PaxSinica_V1: Rich in detail with fair skin tones, blending both Eastern and Western aesthetic sensibilities.

    Moss-PaxSinica_V2: Rich in detail with a graceful, ethereal quality, more closely aligned with East Asian aesthetic preferences.

    Recommended Parameters:

    Any combination at resolutions up to 2160x2880.

    It is highly recommended to use the ComfyUI-Spectrum-MOSS custom node along with comfy-kitchen-attention to optimize your workflow.

    Avoid using any "acceleration" LoRAs.

    Sampler: er_sde

    Scheduler: simple

    Sampling Steps: 30–40

    CFG Scale: 3–4

    Recommended for use with the following two Text Encoders:

    qwen3.5_9b_heretic_v2_int8_hq.safetensors

    qwen3.5_9b_ultra_uncensored_heretic_int8_convrot.safetensors

    #+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-#

    Description

    FAQ

    Checkpoint
    MOSS Image

    Details

    Downloads
    32
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/8/2026
    Updated
    11/10/2026
    Deleted
    -

    Files

    Moss-PaxSinica_V1_int8_convrot.safetensors