CivArchive
    Oilecule - Anima v1.0
    NSFW
    Preview 137432797
    Preview 137432796
    Preview 137432799
    Preview 137432800
    Preview 137432801
    Preview 137432802

    Oilecule

    – Hey, AI Oil Painting Squad, Assemble! Give Your Anime Girls a Splash of Pigment Vibes~

    嘿，AI 油畫黨集合！讓你的動漫妹子沾點顏料味兒~

    Sigh, I gotta admit: I’m a fake-serious oil painting enthusiast.

    哎呀，我得承認：我是个假正经的油画爱好者。

    Oilecule is my weekend mischief product—using Illustrious as the base, I poured in a bunch of oil painting inspirations.

    Oilecule 就是我周末的胡闹产物——拿 Illustrious 当底子，灌了堆油画灵感进去。

    It’s not some rigid “master copy machine,” but it makes your characters look like they just jumped off the canvas: a bit of brushstroke roughness, some ambiguous lighting, and that illusion of “hey, I can smell the turpentine in this pic.”

    不是那种死板的「大师复制机」，而是让你的角色看起来像刚从画布上蹦下来：有点笔触的粗糙，有光影的暧昧，还有那股「哎，这图我能闻到松节油」的错觉。

    What’s So Addictive About It?

    啥玩意兒這麼帶感？

    - Pigment Mood: Not obsessing over “thick impasto,” but you’ll spot those little brush marks, colors lazily blending in layers. Anime faces blend seamlessly—eyes sparkling, hair silky, yet with that oily “touchable” texture.

    - 顏料的調調： 不是硬抠「厚塗」，但你會看到刷子留下的小痕跡、顏色層層暈開的懶散感。動漫臉蛋融得天衣無縫——眼睛亮晶晶，頭髮絲滑，卻帶點油潤的「摸得到」質地。

    - Light Tricks: I love those side-light shadow games; characters look like whispering by candlelight, backgrounds? Misty forests or sunset glows, switch with your mood.

    - 光線小把戲： 愛死了那種側光玩陰影的把戲，人物像在燭火邊低語，背景呢？森林霧氣繚繞還是夕陽餘暉，隨你心情一鍵變。

    - Girl Specials: Skin reflecting light so naturally, curves soft like an invitation for a second glance (but don’t get too immersed). Multi-character scenes hold steady, no collapse even in twisted poses.

    - 妹子專屬： 皮膚反射光那叫一個自然，曲線柔柔的，像在邀請你多瞄兩眼（但別太入戲啊）。多人場景也穩，姿勢再扭也不崩。

    - Artist Cameos: Stuffed with shadows of 40-ish old masters by [whowho], oil vibes does it), from Monet’s dreamy flower fields to Van Gogh’s manic swirls. Full menu? Dive into the example images to explore!

    - 藝術家客串： 塞了 40 來個老傢伙的影子by [誰誰誰], oil vibes 就行），從莫內的夢幻花田到梵高的瘋癲漩渦。完整菜單？例圖中去探索吧！

    - No Fuss: HD straight out (1536x1536 no sweat), no watermark ghosts, toss a prompt and it hits—tested till 3 AM for this ease.

    - 不折騰： 高清直出（1536x1536 隨便怼），無水印鬼影，提示一扔就中——我測試到凌晨三點，就為這省心。

    How I Poke Around Daily

    我平時怎麼戳的

    Resolution: 1024x1536 (vertical screen addicts’ fave) / 832x1216 / other SDXL standards.

    解析度：1024x1536（豎屏控的最愛）/ 832x1216 / 其他 SDXL 標配。

    Steps: 25-35 (28 hits the spot, just right).

    Steps: 25-35（28 夠味兒，不多不少）。

    Sampler: Euler a (for artistic freedom) or DPM++ 2M.

    Sampler: Euler a（給點藝術自由）或 DPM++ 2M。

    CFG: 5-7 (6 steady like my cat).

    CFG: 5-7（6 穩如我家貓）。

    Clip Skip 2, Scheduler Normal.

    Clip Skip 2，Scheduler Normal。

    Positive Lazy Line:

    masterpiece,best quality,amazing quality,very awa,absurdres,newest,very aesthetic,depth of field,highres,

    E.g.: oil vibes, 1girl, blue-haired witch, rainy street at dusk, mysterious smile, warm lamp glow

    正向懶人句：

    masterpiece,best quality,amazing quality,very awa,absurdres,newest,very aesthetic,depth of field,highres,

    比如oil vibes, 1girl, blue-haired witch, rainy street at dusk, mysterious smile, warm lamp glow

    Negative Pitfall Shield:

    worst quality,bad_quality,normal quality,lowres,anatomical nonsense,bad anatomy,anatomical nonsense,interlocked fingers,extra fingers,watermark,simple background,transparent,bad_feet,bad_hands,logo,text,bad_anatomy,signature,face backlighting,(worst quality, bad quality:1.2),jpeg artifacts,censored,extra digit,ugly,deformed anatomy,bad proportions,

    Want wilder? Toss rough strokes. Want gentle? Add soft blends. Hand/eye glitches? ADetailer quick fix.

    負向防坑：

    blurry mess, flat as board, digital plastic, bad hands, watermark nonsense, extra limbs

    想野一點？扔 rough strokes。想溫柔？加 soft blends。手眼小毛病？ADetailer 秒救。

    Example image metadata? Drag it into ComfyUI, copy my crimes~

    例圖的元數據？直接拽 ComfyUI 裡，複製我的罪證去~

    Anima

    Below are some relevant generation settings/data for your reference:

    Works at resolutions between 512^2 and 1536^2 pixels

    • Steps: 20–35 (Best around 30 )

    • CFG Scale: 3-5 (Best around 4 )

    • Sampler: er_sde (or Euler )

    • CLIP Skip: 2

    • VAE: Already integrated

    • Restore Faces: Off

    • Adetailer: On (for clean facial detail)

    ✅ Positive Prompt (Universal & Stable)

    masterpiece, best quality, amazing quality, absurdres,

    🚫 Negative Prompt (Recommended for Clean Results)

    worst quality, low quality, early, old, score_1, score_2, score_3, cartoon, graphic, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, bad anatomy, bad hands, missing fingers, extra fingers, extra digits, fewer digits, cropped, very displeasing, artist name, blurry, jpeg artifacts, lowres, censor

    High-Res

    • Base Resolution: 1024×1024 or larger

    • Highres Fix: On

    • Upscale Factor: 1.5×

    • Hires Steps: 10

    • CFG Scale: 5.5

    • Upscaler: 4x-AnimeSharp

    • Denoising Strength: 0.8

    And here is my personal Anima T2I (Text-to-Image) workflow:

    Reed Anima T2I Workflow

    As you can see in the v10 showcase, I have provided a side-by-side image comparison. The odd-numbered images are raw text-to-image outputs straight from the model, while the even-numbered ones are the upscaled results using 4x-AnimeSharp. I believe this will be super helpful for everyone! All images use the exact same seed and negative prompt so you can see the true, unedited performance. Based on my actual testing, the raw generation (direct output) actually retains more fine details.

    Form Anima Base

    Casual Chat: Why I Messed With This

    閒扯兩句：為啥我折騰這個

    Truth be told, I’m just a “wanna paint but too lazy to death” guy.

    說白了，我就是個「想畫畫卻懶死」的傢伙。

    Scribble notebooks piled into mountains as a kid, grown-up AI saves the day—but those glossy pics always leave me hollow.

    從小塗鴉本堆成山，長大後 AI 救場——但那些光鮮圖總讓我空虛。

    Oilecule is my little protest: make machines learn the “happy accidents” of handmade.

    Oilecule 是我的小抗議：讓機器學學手工的「意外驚喜」。

    Discord drop by? https://discord.gg/52q6zuxcr2 (vent, show off, snag menu).

    Discord 串門？ https://discord.gg/52q6zuxcr2 （吐槽、曬貨、搶菜單）。

    Keep it alive? Buy me a coffee / Buzz (power bill pigments all on it, thanks!).

    續命？Buy me a coffee / Buzz（電費顏料全靠它，謝啦！）。

    Little Rules (No BS):

    Play by Illustrious open-source. Pics are yours to own, pot’s on you. Share for fun, commercial/remix hit me up.

    小規矩（不囉嗦）：

    照 Illustrious 開源玩。圖是你生的，鍋你背。分享樂呵，商業/改版找我聊。

    Come on, poke one—your next wallpaper might be hiding in the pigment pile. 😏🖌️

    來，試戳一張——說不定你的下一個壁紙，就藏在顏料堆裡。😏🖌️

    Description

    FAQ

    Comments (2)

    lamfong223648Jul 23, 2026
    CivitAI

    想問一下大大Anima如何clip skip 2?小弟每次加clip set last layer -2 都出錯，問Al它説Anima只有本體model而沒有clip model跟vae model。

    ReedVaughn
    Author
    Jul 23, 2026

    本模型的详细内容介绍还没有更新，具体参数可以参考 Read XXX Anima 的介绍 本模型下的介绍仅供参考SDXL 模型

    Checkpoint
    Anima
    by ReedVaughn
    Download (Beta) View on CivitAI
    anime
    3d
    base model
    2d

    Details

    Downloads
    608
    Platform
    CivitAI
    Platform Status
    Available
    Created
    7/22/2026
    Updated
    8/13/2026
    Deleted
    -

    Files

    qwen_image_vae.safetensors

    Size:
    242.05 MB
    SHA256:
    a70580f0213e67967ee9c95f05bb400e8fb08307e017a924bf3441223e023d1f

    Mirrors

    HuggingFace (220 mirrors)
    CivitAI (166 mirrors)
    TensorFiles (1 mirrors)
    ModelScope (3 mirrors)
    ModelScope CN (5 mirrors)

    oilecule_animaV10_txt.safetensors

    Size:
    1.11 GB
    SHA256:
    cd2a512003e2f9f3cd3c32a9c3573f820bb28c940f73c57b1ddaa983d9223eba

    Mirrors

    HuggingFace (103 mirrors)
    CivitAI (151 mirrors)
    ModelScope (1 mirrors)
    ModelScope CN (2 mirrors)

    oilecule_animaV10.safetensors

    Size:
    3.90 GB
    SHA256:
    1ca326ae64c63460ad12f894458a48806ab4a6f7c373acfefbb1a0e9e2a9c829

    Mirrors

    CivitAI (1 mirrors)