CivArchive
    Moss Image - Better Nude - v2.0
    NSFW
    Preview 146612556
    Preview 146538698
    Preview 146589540
    Preview 146591906
    Preview 146589987
    Preview 146589545
    Preview 146538719
    Preview 146538722
    Preview 146538723
    Preview 146540111
    Preview 146540550
    Preview 146540755
    Preview 146545801
    Preview 146541795
    Preview 146542278
    Preview 146545099
    Preview 146539574
    Preview 146542648
    Preview 146542651
    Preview 146544559

    V2.0 added more sex scenes
    V2.0 face + pussy merged moss_face_regface_v2.safetensors to V2.0


    workflow.json attached
    Check this post (https://civarchive.com/posts/31908504) for sex scenes prompts.

    This LoRA was trained on MOSS-Image-Base and can also be used with MOSS-Image-SFT, but I find that the base model produces better results.

    Lora strength: 1.0

    Steps: 25 or 50
    CFG: 3 or 4
    Sampler: euler or er_sde
    Scheduler: Simple or Beta
    Resolution: 1024x1536 or 1536x1024

    Only Asian female in the dataset!

    pussy lora for MOSS image

    Description

    v2.0 adds more sex scenes

    FAQ

    Comments (5)

    Samada01Oct 23, 2026· 1 reaction
    CivitAI

    这是我见过最好原RAW风lora了。牛逼,

    AIai88Oct 24, 2026
    CivitAI

    Is moss_face_regface_v2.safetensors a LoRA feature for facial features? Is it your own, or is there a download link?

    salttaro
    Author
    Oct 24, 2026· 1 reaction

    It's just a small face LoRA I trained. I swapped all the faces in the dataset to the default face of Z-Image. V1 overfit to that face, so in V2 I decreased the loss in the face area to 0.3, hoping it would be more compatible with other face LoRAs. So moss_face_regface_v2.safetensors isn't critical, just there to test this LoRA's compatibility with other face LoRAs.

    ZiYao424Oct 25, 2026
    CivitAI

    Hello, I'm testing on the RH platform. If I load 3-4 LoRa simultaneously, it causes the workflow generation to fail. I tested for 3 hours and confirmed that it's this LoRa causing the issue, but it works fine when used alone. I'm using 8 steps.

    salttaro
    Author
    Oct 25, 2026

    Thanks for testing! Could you share the error message?

    One possible cause: this LoRA trains MOSS's fused qkv layer, while some other MOSS LoRAs (e.g. the 8-step ones) use separate q/k/v. Recent ComfyUI handles mixing them fine (works when I stack them locally), but older versions or some loader nodes may not. The error text, plus RH's ComfyUI version and LoRA loader node, would help confirm.

    LORA
    MOSS Image

    Details

    Downloads
    5,569
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/23/2026
    Updated
    10/27/2026
    Deleted
    -

    Files

    moss_pussy_lora_v2.0.safetensors

    workflow.json

    Mirrors

    CivitAI (2 mirrors)