CivArchive
    anima-matte v1.1 - v0.1
    NSFW
    Preview 129589356
    Preview 129589359
    Preview 129589375
    Preview 130104760
    Preview 130104774

    anima-matte v1.1

    This is a LoRA slider that reduces skin gloss for a matte finish, or conversely, increases gloss.

    In short, it removes or increases highlights in the skin.

    This time, in accordance with Anima Base 1.0, we changed the training images, disabled TextEncoder training, and changed the resolution to 1024x1024.

    For a matte finish, a LoRA weight of around +0.45 is recommended.

    For a shiny finish, a LoRA weight of around -0.45 is recommended.

    Start with ±0.45 and then try increasing or decreasing it.

    If the original color scheme lacks highlights, such as in anime, this will have absolutely no effect; in fact, it will cause you to lose the brightness of fine details.

    In this example, we used a training image created by combining a fixed background with the subject, but it appears that the influence of the background cannot be completely eliminated.

    皮膚の光沢を抑えマット化、逆に光沢を増やすスライダーLoRAです。

    要するに皮膚のハイライトを消したり増やしたりします。

    今回はAnima Base 1.0に合せ、教師画像の変更とTextEncoder学習の無効化と解像度1024x1024への変更を行いました。

    マットにしたい場合、LoRAのウェイトは+0.45位が良いと思います。

    テカテカにしたい場合、LoRAのウェイトは-0.45位が良いと思います。

    先ずは±0.45で始め増減してみてください。

    アニメカラー等ハイライトが元から無い場合全く効果は無いです、逆に輝度の高い細かい箇所を失わせてしまいます。

    今回は固定された背景と被写体を合成した教師画像を使用しましたが、背景に対する影響は打ち消せない様子です。

    Notes:

    Please note that there will be some variation in style and composition; even with the same seed, the composition will not be exactly the same.

    Since v1.0 was trained with Anima base Preview3, I retrained it with Anima base v1.0.

    A side effect of LoRA is that, instead of only altering the highlights and shadows of the subject, the effect extends to the entire image. This is particularly noticeable in flat image styles, but it works reasonably well if you want to control the highlights in semi-realistic image styles.

    画風や構図に多少なりとも変化が加わります、同一シードであっても全く同一の構図にはならない点に注意して下さい。

    v1.0はAnima base Preview3で学習してしまった為、Anima base v1.0で学習し直しました。

    このLoRAの副作用として、被写体に限ったハイライト・シャドゥの変更を行うのではなく、効果が画像全体に及びます。特にフラットな画風では顕著になる様子です、セミリアル系画風のハイライトをコントロールしたい場合はそれなにり機能します。

    Description

    FAQ

    Comments (3)

    mac2492May 8, 2026
    CivitAI

    If I'm not mistaken, this seems to affect all highlights and shadows across the image (and not just skin glossiness). It's pretty interesting for realistic/semi-realistic gens but kinda just confuses the model with flat styles.

    OMCCOM
    Author
    May 8, 2026· 1 reaction
    Indeed, this LoRA (AI) cannot recognize only the subject and suppress highlights. If you have any ideas for improvement through learning, I will try them out.

    mac2492May 8, 2026· 1 reaction

    I think it's actually okay for the LoRA to affect the entire image. In this case you could probably just change the description haha

    I'm not familiar with training at all, sadly. The only thing I can think would be to use images with different styles + backgrounds then create variations with glossy/flat subjects with the background untouched. I don't know if this would translate to the model being able to isolate the subject, and I think it'd be too much work when users can just use the LoRA with masking / regional prompting if they want to isolate the effect.

    LORA
    Anima

    Details

    Downloads
    156
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/4/2026
    Updated
    5/25/2026
    Deleted
    -

    Files

    anima-matte-v01.safetensors

    Mirrors

    CivitAI (1 mirrors)