CivArchive
    Reactive - V1
    NSFW
    Preview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefinedPreview undefined

    ⚙️ Recommended Settings

    This is the best setting for realistic faces, skin, details of clothes, details in background, etc.
    CFG Scale 4-8
    Clip Skip 1
    Sampler DPM++ 3M SDE Exponential
    Sampling Steps 35-45 for low quality results, 45-75 for mid quality. You can also go higher for more details.
    VAE None // or vae-ft-mse-840000-ema-pruned if you really need one.

    ✍️ Recommended workflow

    • Set picture size to 576x1024 (for portraits) or 1024x576 (for landscapes). It's 16:9 aspect ratio. Don't go above 1024px!

    • Use negative embeddings or good negative prompt. You can find popular embeddings here.

    • Use 8x_NMKD-Superscale_150000_G or 4x-Ultrasharp for Upscale (strenght 0.2-0.4)

    • (optional) Put nationality into positive prompt, for example (Irish girl:1.3) or (Malaysian woman:1.3) to modify face shape and skin tone. You can read more about this in my article.

    • (optional) Put age into positive prompt. Be specific. Use keywords like: XX years old, girl, woman, old lady, elderly woman.
      Combining these words will give you the best results. Examples:
      photography of German girl, 20 years old
      photography of Italian woman, 40 years old
      photography of Spanish woman, very old lady, elderly woman
      More info here.

    • Do not use Celebrity / face morphing Textual Inversions because in most cases it will make your skin unnatural, pale. Better find good Lora and use it instead.

    • Do not use Style of artist X or ...by artist Name Surname prompts if you expect realistic skin.

    Recommended Lora's

    • Weight Slider (use negative values for skinny girls and positive for thick girls)

    • Shiny Oiled Skin (use Only for nudes, otherwise it can make your faces ugly)

    • LowRA (use it for professional studio photos, for non-studio use about 0.2-0.3)

    • FujiFilm Instax Polaroid (nice Lora for old photography effects)

    🧾 Recommended Embeddings (Textual Inversions)

    👎 Known weak points of this model

    • Sometimes nipples are trying to break through the clothing 😢

    • Sometimes straps in bras and panties are missing 😠

    • This model is not so flexible with Nationalities as my another model AddictiveFuture Realistic SemiAsian 🙁

    // all models and images were creted on Vast.ai servers (using RTX4090)
    // most of images for this model were generted with ADetailer (face_yolov8n)

    Description

    Checkpoint
    SD 1.5

    Details

    Downloads
    415
    Platform
    CivitAI
    Platform Status
    Available
    Created
    2/24/2024
    Updated
    9/30/2025
    Deleted
    -

    Files

    reactive_v1.safetensors

    Mirrors

    CivitAI (1 mirrors)
    Other Platforms (TensorArt, SeaArt, etc.) (1 mirrors)