CivArchive
    Preview 117501574
    Preview 117501573
    Preview 117504713
    Preview 117627311
    Preview 117505544
    Preview 117518449
    Preview 117519988
    Preview 117521392
    Preview 117534986
    Preview 117555299
    Preview 117616444
    Preview 118041065
    Preview 118101437
    Preview 118110761

    🦑 Ikastrious 🦑

    クリアな線と発色が特徴の、高品質な汎用アニメモデル

    A high-quality, versatile anime-style model featuring clear linework and vibrant colors.

    --prompt--
    (explicit|general), masterpiece, best quality, anime coloring (if you need)
    --negative--
    lowres, worst quality, bad quality, worst detail

    Recommend Settings

    • Sampling method & Schedule type: Euler a & automatic

    • Sampling steps: 10~40

    • CFG Scale: 4~7

    • Resolution: 1024x1024, 896x1152, 832x1216 / 1216x1216, 1024x1344,

    My Favorite Settings:
    
    Sampler: Euler a, Schedule type: Automatic, Steps: 27, CFG scale: 4.5
    Hires upscaler: 4x-UltraSharpV2_Lite (fast!) or 4x-AnimeSharp
    Hires steps: 10, Denoising strength: 0.2, Hires upscale: 1.5, Hires CFG Scale: 1

    Description

    マージするモデルを見直しつつも、鮮やかな色合と頭身は維持しました
    While revisiting the model merge, I preserved the vivid color palette and the original proportions.

    FAQ

    Comments (2)

    antarekJan 15, 2026· 8 reactions
    CivitAI

    I tend to find that most of the time DPM2 a + normal produces way better results than Euler a + automatic/normal/simple/karras. Also DPM2 a is a sampler that works extremely well at high steps (70+) if you want to generate extremely detailed pics. It also works way better with anime themed content than DPM++/ DPM++ 2M/DPM++ 3M samplers.

    giko
    Author
    Jan 15, 2026

    Your settings take time but yield very accurate results! I'll keep that in mind.

    Checkpoint
    Illustrious
    by giko

    Details

    Downloads
    1,250
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/15/2026
    Updated
    7/21/2026
    Deleted
    -

    Files

    ikastrious_v200.safetensors