CivArchive
    Hipoly 3D Model LoRA - v2.0
    NSFW
    Preview 485327
    Preview 485328
    Preview 485329
    Preview 485330
    Preview 485331
    Preview 485332
    Preview 485333
    Preview 485334
    Preview 485335
    Preview 485337
    Preview 485339
    Preview 485341

    (日本語は後半にあります)

    High-poly 3D Model LoRA

    This is a LoRA trained on high-polygon 3D model images.

    It can provide clean, high-resolution skin and hair materials, as well as detailed clothing elements.

    For those interested, I have compiled the technical insights gained during the training of ver.2 in the following article.

    https://note.com/takumi__ncr/n/n21016c358ea5

    Changes in 2.0

    • Increased the number of training images

    • Increased the training resolution (ver.1: 768, ver.2: 896)

    • Revised the tagging

      • Removed the trigger word as well

    • Reviewed the training parameters

    Improvements in 2.0

    • It can reproduce a more 3D-like texture and stereoscopi effect than ver.1

    • LoRA can be applied without a trigger word

    • The style can be controlled using 3d and realistic tags

      • When added to Positive Prompt, it enhances the 3D feel

      • When added to Negative Prompt, it adds details such as clothing while maintaining the model's art style

    • The issue of eroding the model's art style with tags such as intricate, detailed has been alleviated

    Additional information

    The effect of LoRA was confirmed using the following my custom merged models:

    High-poly 3D Model LoRA

    ハイポリゴン 3D モデル画像を学習した LoRA です。

    3D らしいクリーンで高精細な肌・髪の質感や、衣服のディテールを与えることができます。

    ver.2 の学習時に得られた技術的知見をまとめましたので、興味がある方は次の記事をご覧ください。

    https://note.com/takumi__ncr/n/n2fb9d265ffa9

    2.0 での変更点

    • 教師画像を増やしました

    • 学習解像度を向上しました (ver.1: 768, ver.2: 896)

    • タグ付けを見直しました

      • トリガーワードも削除しました

    • 学習パラメータを見直しました

    2.0 での改善点

    • ver.1 よりも 3D らしいテクスチャ感、奥行き感が再現できます

    • トリガーワードなしで LoRA を適用することが可能です

    • 3d, realistic タグを使用することで絵柄のコントロールが可能です

      • Positive Prompt に入れた場合、より 3D 感を強化します

      • Negative Prompt に入れた場合、モデルの絵柄を保ちつつ、衣装などのディテールを追加することができます

    • intricate, detailed などのタグによってモデルの絵柄を侵食してしまう問題が緩和されました

    その他

    LoRA の動作確認は以下の自作マージモデルを使用して行いました。

    Description

    • Increased the number of training images

    • Increased the training resolution (ver.1: 768, ver.2: 896)

    • Revised the tagging

      • Removed the trigger word as well

    • Reviewed the training parameters

    FAQ

    Comments (23)

    HosiokaApr 13, 2023· 1 reaction
    CivitAI

    If you're down to provide me the dataset ill train you a LoCon version with a higher resolution. you're using far too high of a DIM

    NotAPersonJustACatApr 13, 2023· 1 reaction
    CivitAI

    Thanks for the article.

    qs7Apr 13, 2023
    CivitAI

    So I can just put this with other loras?

    Tian_DGGApr 22, 2023

    yes

    novakardApr 13, 2023· 1 reaction
    CivitAI

    Those notes you put together were a fascinating read, thank you so much for doing so!!

    ailorachanApr 15, 2023· 1 reaction
    CivitAI

    it seems like v20 is really strong when mixing with other loras. v10 still considered more on hair and facial definitions however v20 changes quite a lot of the other loras used when mixing

    WarTornApr 17, 2023· 1 reaction
    CivitAI

    one of the best loras out there.

    65473Apr 21, 2023· 1 reaction
    CivitAI

    What's the trigger word?

    Tian_DGGApr 22, 2023· 3 reactions

    1girl

    Ansen_WuApr 26, 2023

    @Tian_DGG pu~~Anything goes 1girl, hhh

    KevinKang_NApr 22, 2023· 3 reactions
    CivitAI

    2.0 can make quite big influence when used with other loras, it is not suitable for making cosplay picture. 1.0 with lower weight is quite more good to do the same thing.

    bsndMay 5, 2023· 1 reaction
    CivitAI

    The notes you did up are really interesting & a great read, thank you!

    na1kedartMay 27, 2023
    CivitAI

    i tried your lora and model but colors seem very washed out ; is there something i'm doing wrong?

    9mmheater919May 29, 2023

    You need to add a VAE

    na1kedartJun 1, 2023

    mhm i have done that already

    luizespalhalixoJun 8, 2023

    @aiarts me too bro. Can you link the VAE and how to you solve it? Thanks!

    hugsyyy008Jun 30, 2023
    CivitAI

    something not right <lora:hipoly3DModelLora_v20:0.6> actually is <lora:hiqcg_body_epoch:0.6> PLS FIX!

    shunwakae167Jul 10, 2023
    CivitAI

    かわいいですね。じぶんでつくるのは、まだむずかしいので、みるがわにいます

    RayShepardJul 16, 2023
    CivitAI

    <lora:hipoly3DModelLora_v20:0.7>,<lora:hipoly_3dcg_v7-epoch:1>,before one is the samply used,while the author provieded the another one.and i can read a message from controller:Couldn't find Lora with name hipoly3DModelLora_v20

    求V20的Lora啊啊啊啊啊T^T

    NukeDayAug 29, 2023

    You have hipoly_3dcg_v7-epoch, so delete hipoly3DModelLora_v20 from the prompt. Maybe training the Lora uses another name, but when uploading it's not available and the authors change it.

    joe_stapfNov 6, 2023
    CivitAI

    Tried this two phrases in your prompt with NMKD:

    withLora(hiqcg_body_epoch)0.6

    realistic

    and wow, what a difference! :)

    hascapJun 1, 2024· 1 reaction
    CivitAI

    I wish sdxl

    DjojiSep 5, 2024
    CivitAI

    Your Lora and his polygon model can have a good effect of many 3d Rendition, ty for creating it

    LORA
    SD 1.5

    Details

    Downloads
    96,533
    Platform
    CivitAI
    Platform Status
    Available
    Created
    4/13/2023
    Updated
    6/11/2026
    Deleted
    -

    Files

    hipoly_3dcg_v7-epoch-000012.safetensors

    Mirrors

    HuggingFace (65 mirrors)

    Available On (3 platforms)

    Same model published on other platforms. May have additional downloads or version variants.