CivArchive
    Face Sitting V2 - Couple & Interaction Pose (SD 1.5 LoRA) - v1.0
    NSFW
    Preview 145618426
    Preview 145618415
    Preview 145618421
    Preview 145618418
    Preview 145618424

    Model Name: Face Sitting V2 - Couple & Interaction Pose (SD 1.5 LoRA)

    Type: LORA

    Base Model: SD 1.5

    Trigger Words: face sitting

    Trained Tags: 1woman, 1man, 2people, woman sitting on man face, man face underneath, looking at viewer, looking back, from behind, back view

    Recommended Weight (Strength): 0.85 - 0.95 (Tested best at 0.9)

    Clip Skip: 1

    # 🍑 Face Sitting V2 - Dynamic Couple Pose (SD 1.5 LoRA)

    Face Sitting V2 is a refined pose and interaction LoRA for SD 1.5, specifically designed to overcome the common multi-subject generation issues of coupled poses. 

    Trained with precise individual image captions and directional tags, V2 ensures the male partner's head and neck are anatomically grounded beneath, while giving you complete directional control over the female partner (looking back at viewer or facing away).

    ---

    ### 🌟 Key Features & Improvements:

    * Stable 2-Person Interaction: Locks the composition so the male partner doesn't disappear or get warped.

    * Directional Freedom: Easily switch between Looking at Viewer (over the shoulder) and Back View.

    * High Outfit Flexibility: Works seamlessly with casual outfits, skirts, bodysuits, lingerie, or swimwear.

    * Trained with High-Precision (Dim 32 / Alpha 16): Preserves crisp curves and natural skin textures.

    ---

    ### ⚙️ Recommended Generation Settings (Verified Preset):

    * **Base Checkpoint:** [PuppetCrazy Mix V1.5 / V1.7.1](ใส่ลิงก์โมเดลของพี่ตรงนี้) (Primary Recommendation for best aesthetic skin & anatomy)

    * LoRA Weight: 0.85 ~ 0.95 (Sweet spot: 0.9)

    * Sampling Method: Euler a (or DPM++ 2M Karras)

    * Steps: 25

    * CFG Scale: 6.0 ~ 7.0

    * Resolution: 512x640 or 512x768

    * Clip Skip: 1

    ---

    ### 🎯 Pro Tip: ADetailer Configuration (Essential):

    To achieve clean facial expressions without the character unintentionally closing her eyes:

    * Detector: face_yolov8n.pt

    * Inpaint Denoising Strength: 0.25

    * ADetailer Prompt: (Leave empty)

    * ADetailer Negative Prompt: (1woman close eyes:1.3),

    ---

    ### 💬 Sample Prompt Templates:


    #### 1. Looking at Viewer (Turned Head / Eye Contact):

    Positive Prompt:

    ```text

    (masterpiece, best quality:1.2), realistic photo, <lora:face_sitting_v2-000007:0.9>, face sitting, 1woman, 1man, 2people, looking at viewer, looking back, woman sitting on man face, man face underneath, soft indoor lighting, sharp focus

    (masterpiece, best quality:1.2), realistic photo, <lora:face_sitting_v2-000007:0.9>, face sitting, 1woman, 1man, 2people, from behind, back view, facing away, woman sitting on man face, man face underneath, soft indoor lighting, sharp focus

    (worst quality, low quality:1.4), (bad anatomy:1.2), (bad hands:1.2), missing fingers, extra fingers, fused fingers, deformed, distorted, blurry, (bad proportions:1.2), watermark, doll, anime, young, teenager, round face, chubby face, cute, smooth plastic skin, 2girls, duplicate, 2 pussy, 2 anus, 3 hands, 3 legs, badhandv4, ng_deepnegative_v1_75t, blurry

    Description

    LORA
    SD 1.5

    Details

    Downloads
    26
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/14/2026
    Updated
    10/20/2026
    Deleted
    -
    Trigger Words:
    face sitting

    Files

    face_sitting_v2-000007.safetensors

    Mirrors