CivArchive
    Manmaru mix - v2.0
    Preview 2660042
    Preview 1525678
    Preview 1525688
    Preview 1702297
    Preview 1525786
    Preview 1525795
    Preview 1537483
    Preview 1537657
    Preview 1539850
    Preview 1525919
    Preview 1526058
    Preview 1526169
    Preview 1540038

    I don't speak English so I'm translating at DeepL. Let me know if the English is weird.

    This model is based on the Thumbelina v2.0 Model character sketch,
    which was adjusted to make it look less hand-drawn and to make the background appear clearer.

    このモデルはThumbelina v2.0 Modelのキャラデをもとに、
    手描き感を薄くし、背景がくっきり出るように調整しました。

    Update History / 更新履歴

    • V2
      The background is clearer than in V1. Also, the hands should be just a little bit better...
      After using it myself for a while, I noticed that the face of the output is cuter than V1. That's my subjective opinion.
      背景がV1よりよりくっきりしました。あと、ほんのちょっとだけ、手の描写がましになっているはず・・・
      しばらく自分で使ってみて、顔がV1よりかわいくなっていることに気づきました。私の主観ですが。

    • V3
      The hand rendering is more decent.
      It is still not as good as other models, but it is much better.
      Also, the background is clearer and the overall composition is less broken.
      手の描写がよりまともになっています。それでもまだほかのモデルに比べたら
      悪いですが、だいぶましになりました。
      また、背景がよりはっきりとなり、全体の構図も壊れることが少なくなっています。

    Merging Method / マージ方法

    V1

    1. A:Thumbelina_v20.safetensors
      B:nostalgiaMix_nostalgia.safetensors
      Sum Merge : -0.6 (THIS IS "minus"!! マイナスです!!)
       ->tmp.safetensors

    2. A:flat2DAnimerge_v20.safetensors
      B:tmp.safetensors
      C:featurelessMix_v10.safetensors
      MBW:
       alpha:0.25,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5
       beta:0.5,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0

    V2

    1. A:flat2DAnimerge_v20.safetensors
      B:cocotifamix_v20This.safetensors
      Sum Merge : 0.5
       ->tmp.safetensors

    2. A:manmaruMix_v10.safetensors
      B:tmp.safetensors
      MBW:
      alpha:0,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,1,1,1

    V3

    There are so many steps because of trial and error!
    試行錯誤したからめっちゃ段階多いよ!

    1. A:Thumbelina_v20.safetensors
      B:nostalgiaMix_nostalgia.safetensors
      sum merge : -0.4 (this is minas / マイナスです)
      -> tmp01.safetensors

    2. A:tmp01.safetensors
      B:featurelessCuteMix_v10.safetensors
      sum merge : 0.5
      -> tmp02.safetensors

    3. A:tmp02.safetensors
      LoRA merge : 3dSlider_JRPG_v1:0.25
      -> tmp03.safetensors

    4. A:manmaruMix_v20.safetensors
      B:featurelessCuteMix_v10
      sum merge : 0.1
      ->tmp04.safetensors

    5. A:tmp04.safetensors
      LoRA merge : 3dSlider_JRPG_v1:0.25
      ->tmp05.safetensors

    6. A:tmp04.safetensors
      B:tmp05.safetensors
      sum merge : 0.5

     z-vae v2.0 has been merged. / z-vae v2.0をマージ済みです。(thx verxonous86495!)

    Origin of the model name / モデル名の由来

    "manmaru" = "まん丸", means "perfectly round".
    Because the eye shape output by this model is round.

    このモデルで出力される目の形がまん丸なので、"manmaru"。

    Description

    FAQ

    Comments (9)

    SLAPaperJul 12, 2023· 2 reactions
    CivitAI

    How about animals generation? like cat, dog, bird, panda?

    pawapawa
    Author
    Jul 14, 2023

    Doesn't it appear when you type these words in the prompt?

    kusogeJul 14, 2023· 7 reactions
    CivitAI

    One of the best stylized models.

    Works very well with character LORA's

    fakehjx4Aug 5, 2023· 2 reactions
    CivitAI

    このモデルはどのVAEと使えばいいですか?(下手な日本語で質問をしてごめんね(笑))

    pawapawa
    Author
    Aug 5, 2023

    このページのサンプル画像はanything-V3.0.vaeを利用しています。また、vae-ft-mse-840000-ema-pruned.vaeもよく合うと思います。私は新しいVAEを全然試してないので、合うものを見つけた時はここに書いていただけると嬉しいです。

    The sample image on this page uses anything-V3.0.vae. I also think sd_vae-ft-mse-840000-ema-pruned.vae fits well. I haven't tried any of the new VAEs at all, so if you find one that fits, I'd be happy to write about it here.

    cornerbearSep 23, 2023· 11 reactions
    CivitAI

    Will this ever get an update to XL or something similar. I really, really love this checkpoint/model :)

    pawapawa
    Author
    Sep 24, 2023· 7 reactions

    Thank you so much for using the ManmaruMix model! I'm glad to hear that you love it—I'm also a big fan of the design of this model. Since releasing ManmaruMix, I have introduced several new models, but I still find the facial design of this one to be my favorite.

    ManmaruMix was created by combining multiple models designed for SD1.5. As for SDXL, there aren’t as many models available to experiment with designs, so creating a similar model for it is challenging at the moment.

    To be honest, another issue is that my PC's specifications are relatively low, which makes it difficult to satisfactorily merge models for SDXL.

    cornerbearSep 24, 2023· 1 reaction

    @pawapawa Thank you so much for the reply. Haha' if you lived close by you could have used my PC. I hope you get some better hardware at one point, and again, awesome job you do :)

    Kind Regards
    Cornerbear