CivArchive
    MomoiroLoRA - LoRA1
    NSFW
    Preview 9940309

    Expose the two LoRAs used to create MomoiroPony. If you are interested in how to create this model, see the following instructions.

    MomoiroPonyを作成するために使用した2つのLoRAを公開します。

    このモデルの作成方法に興味がある方は以下の説明もご覧ください(日本語版は下にあります)。

    How to Create MomoiroPony v1.4

    Translated by MiraiTranslate:

    MomoiroPony was created by merging the Pony Diffusion V6 XL with two LoRAs. I forgot the merge ratio, but it is about the following ratio.

    LoRA1: lbw=0,0,0,0,0.7,0,1,0.55,1,1,1,1

    LoRA2: lbw=0,0,0,0,0,0,0,0.35,0,0,0,0

    LoRA1 is used to stabilize styles. LoRA2 is used to fine-tune styles.

    LoRA1(nsfw70)

    25 images generated by the SD 1.5 model are used for learning. The images are all 1girl, simple background, completely nude. These images are used for learning to avoid affecting the background and clothing Pony is learning. Check the metadata in the LoRA file for learning parameters. However, it is not a good parameter and should not be referenced.

    LoRA2(sfw1)

    17 images with backgrounds are used for learning. Learning a simple background image tends to simplify the background of the generated image. Use this LoRA to mitigate.


    MomoiroPony v1.4 の作成方法

    MomoiroPonyはPony Diffusion V6 XLと2つのLoRAをマージすることで作成されています。

    マージ比率は忘れてしまいましたが、だいたい以下の比率です。

    LoRA1: lbw=0,0,0,0,0.7,0,1,0.55,1,1,1,1

    LoRA2: lbw=0,0,0,0,0,0,0,0.35,0,0,0,0

    LoRA1はスタイルを固定するために使用します。

    LoRA2はスタイルを微調整するために使用します。

    LoRA1(nsfw70)

    SD1.5モデルで生成した25枚の画像を学習に使用します。

    画像は全て1girl, simple background, completely nudeです。

    Ponyが学習している背景や服装に影響を与えるのを避けるためこのような画像を学習に使用します。

    学習パラメータはLoRAファイルのメタデータを確認してください。

    ただし良いとは言えないパラメータのため参考にはしないでください。

    LoRA2(sfw1)

    背景が含まれた17枚の画像を学習に使用します。

    simple backgroundの画像を学習させると生成画像の背景もシンプルになる傾向があるため、このLoRAを使用して緩和させています。

    Description

    FAQ

    Comments (7)

    fansayApr 16, 2024· 1 reaction
    CivitAI

    Quite a pretty style, rounds the face and eyes. I recommend weight 0.8

    DeathspikeApr 17, 2024· 1 reaction
    CivitAI

    I fell in love with MomoiroPony, then fell in love with HanamomoPony. Now, MomoiroLora dials that love to the extreme. This, in combination with Saturation Tweaker, really allowed me to dial in the perfect style to suit my tastes. Thank you!

    PertaliteMeisterApr 17, 2024· 1 reaction
    CivitAI

    love these lora, but now I became greedy and want hanamomo lora, because bigger eyes and smaller eye gap.

    oritatami_neko
    Author
    Apr 18, 2024

    I uploaded it. Note that this is DoRA, not LoRA.

    https://civitai.com/models/406965/hanamomodora

    Beef_ChanAug 19, 2024· 1 reaction
    CivitAI

    Love this model and how colorful it is. It works really well with all of the loras I make. When getting the dataset for lora2 did you use images that were generated with 1.5 with backgrounds or were they just images you liked the style of?

    oritatami_neko
    Author
    Aug 20, 2024· 1 reaction

    LoRA2 also includes a small number of AI-generated images. It mainly selects bright, saturated images.

    Beef_ChanAug 28, 2024

    @oritatami_neko Thanks for responding! I've trained a couple and gotten some interesting results. I've had good luck with keeping the loras and their weights in the prompt vs. combining them with the checkpoint.

    LORA
    Pony

    Details

    Downloads
    741
    Platform
    CivitAI
    Platform Status
    Available
    Created
    4/15/2024
    Updated
    5/16/2026
    Deleted
    -

    Files

    nsfw70.safetensors

    Mirrors

    CivitAI (1 mirrors)
    TensorFiles (1 mirrors)

    Available On (1 platform)

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