CivArchive
    HyperAnima Plastique - 2.9B Turbo - v1
    NSFW
    Preview 144139720
    Preview 144139723
    Preview 144139736
    Preview 144139742
    Preview 144139725
    Preview 144139721
    Preview 144139729
    Preview 144139727
    Preview 144139731
    Preview 144139732
    Preview 144139733
    Preview 144139738
    Preview 144139741
    Preview 144139744
    Preview 144139737
    Preview 144139739
    Preview 144139726
    Preview 144139718
    Preview 144139719
    Preview 144139730

    About HyperAnima Plastique

    HyperAnima Plastique is a Turbo checkpoint based on Anima 2.9B by Gazingstars123, tuned for a more semi-realistic, polished visual style.

    It favors soft volumetric lighting, smooth glossy skin and rich material detail, sitting between illustration and photography, with faces and bodies that read as real while keeping a clean, retouched, editorial finish.

    It is well suited for portraits, fashion, glamour shots, close-ups and cinematic city scenes, with a consistent look that favors luminous skin, detailed eyes and lips, and fabrics that catch the light.

    HyperAnima Plastique is built for fast generation, producing strong results at very low step counts and low CFG.

    Key Features

    • Strong semi-realistic visual identity

    • Smooth, luminous skin with soft specular highlights

    • Detailed eyes, lashes and glossy lips

    • Rich material rendering for satin, velvet, fur, leather, metal and jewelry

    • Soft studio lighting and warm cinematic ambient light

    • Handles a wide range of body types and facial features

    • Coherent real-world backgrounds such as streets, cafés, architecture and waterfronts

    • Supports TAG-style prompting

    • Supports natural-language prompting

    • Designed for very low step counts

    Sampler: ER SDE
    Scheduler: BETA
    Steps: 8–12
    CFG: 1–2

    Sweet spot: 10 steps / CFG 1.5

    Higher step counts are generally unnecessary.

    Recommended positive prefix:
    masterpiece, best quality, score_7, safe, realistic, photorealistic,

    Recommended negative:
    worst quality, low quality, score_1, score_2, score_3, artist name, blurry, jpeg artifacts, chromatic aberration

    Prompting

    HyperAnima Plastique supports both TAG-style prompts and natural language, just like other Anima-based models.

    Both approaches can also be mixed in the same prompt.

    For best results, use TAGs for character traits, appearance, outfit and framing, while using natural language to guide lighting, mood, location and time of day.

    The model responds especially well to prompts that clearly describe the subject, pose, expression, framing and background, especially when paired with strong lighting cues such as soft studio light, golden hour or dusk with warm street lamps, and material cues such as satin dress, velvet gown or faux fur coat.

    Description

    FAQ

    Comments (4)

    TripticnonamedOct 3, 2026· 1 reaction
    CivitAI

    Seems this model doesn't react to any ANI Loras.

    tripledamn
    Author
    Oct 4, 2026

    That’s somewhat expected. Plastique has a strong semi-realistic bias, so style/art LoRAs may have limited effect.

    Character LoRAs, sliders and other non-style LoRAs should still work, though you may need to adjust the weight.

    TripticnonamedOct 4, 2026

    By far this is the best model I've tried in, with a realistic approach on Anima, but i'll ended up using your turbo model, since it adapts better to loras. Keep the good work, Triple.

    tripledamn
    Author
    Oct 4, 2026

    @Tripticnonamed thank you so much, I'll try to release a 2.9B version of Turbo model soon ;)

    Checkpoint
    Anima
    by tripledamn
    Download (Beta) View on CivitAI
    anime
    base model
    semirealistic

    Details

    Downloads
    630
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/29/2026
    Updated
    11/15/2026
    Deleted
    -

    Files

    hyperanimaplastique_v1_turbo29b.safetensors

    Size:
    5.44 GB
    SHA256:
    f0fc8a995cee2504c9544e57da6363e9525e10f911daedecc1a2d6121ce9403d

    Mirrors

    CivitAI (1 mirrors)

    qwen_3_06b_base.safetensors

    Size:
    1.11 GB
    SHA256:
    cd2a512003e2f9f3cd3c32a9c3573f820bb28c940f73c57b1ddaa983d9223eba

    Mirrors

    HuggingFace (148 mirrors)
    CivitAI (277 mirrors)
    ModelScope (1 mirrors)
    ModelScope CN (2 mirrors)