CivArchive
    IU - IU V3.5
    Preview undefined
    Preview undefined
    Preview undefined

    Description

    Merged V3.0 and V2.0 as adviced.

    FAQ

    Comments (38)

    modestThinkerMar 5, 2023
    CivitAI

    I'm using the v1-5-pruned-emaonly.ckpt as the base model and apply the same promot in the discussion.But the outcome pictures are very strange.I need help

    Ithlinni
    Author
    Mar 5, 2023

    use chilloutmix

    kylocatMar 5, 2023
    CivitAI

    may I ask what is the ratio/weight used for V2 and V3 when you merge them into V3.5? Your merged model has way different results than my own merge

    Ithlinni
    Author
    Mar 5, 2023

    0.8 of v3 and 0.2 of v2. Get that idea from the review of jetek72554

    dugemwaeMar 6, 2023
    CivitAI

    hi, mind sharing how to do lora training ? i tried to train lora on kpop star but the result always gets me a bad generated image, like so much noises

    Ithlinni
    Author
    Mar 6, 2023

    Try a lower learning rate or test the result of other epochs instead of the last one. For this one I used 4e-5 unetlr and 5e-6 txlr with batch size 3.

    It's not always good result for me either. Just keep trying.

    dugemwaeMar 6, 2023

    @Ithlinni ah okay, thanks for the quick reply, mind sharing me how many images used and how many times one image is trained do you usually use ? and is it kohya GUI or dreambooth AUTOMATIC1111 ?

    Ithlinni
    Author
    Mar 7, 2023· 4 reactions

    @dugemwae For this one I used 20 images with 20 repeats. And colab of kohya GUI.https://colab.research.google.com/github/Linaqruf/kohya-trainer/blob/main/kohya-LoRA-dreambooth.ipynb#scrollTo=Rt7CKCog_4tm

    404mkMar 14, 2023

    @Ithlinni hi, May i know the suggestion of Training Steps Per Image (Epochs), unetlr and txlr when use dreambooth? Because the traning result is not similar with the target person. Thank You.


    anonyvpn001962Mar 14, 2023
    CivitAI

    why many peopel use SD1.5 instead of 2.0?

    aneroMar 25, 2023· 1 reaction

    because there are some limits on nsfw on the 2.0SD

    faker12Jun 26, 2023

    @anero like what?

    faker12Mar 17, 2023
    CivitAI

    Please think about doing iz*one jo yuri please!

    ghcit12Mar 25, 2023
    CivitAI

    kim hee sun please

    ghcit12Mar 29, 2023
    CivitAI

    kim hee sun please!!!!!!!!!

    username094Mar 31, 2023
    CivitAI

    how do you do loras??

    mouraoMay 24, 2023

    https://civitai.com/models/22530/guide-make-your-own-loras-easy-and-free

    wisesayingstoptenApr 4, 2023· 5 reactions
    CivitAI

    유튜브썸네일로 사용하려고 합니다. AI이미지라 저작권이 없는건 알지만 그래도 메세지 남깁니다 ^^

    ukmi1202242Apr 10, 2023

    오우... ai 생성물 자체가 저작권을 따지지 않는다고는 들었는데 대놓고 아이유 학습한거 올리면 위험한거 아님...?

    howaboutnoApr 5, 2023· 1 reaction
    CivitAI

    These are amazing. I noticed you said 3.5 is 3.0 and 2.0 merged. I was wondering how you go about merging LoRAs?

    jungdaso822Apr 7, 2023· 6 reactions
    CivitAI

    좋은 모델을 만들어주심에 우선 감사드립니다. 유투브 제작의 님의 모델을 사용할 수 있는지 여쭙습니다. 모델을 통해 좋은 이미지를 만드는 연습을 잘 하고 있습니다.

    noname2026Apr 13, 2023
    CivitAI

    Hello, respected author, I really like Kim Na Jung. Can you train Lora?

    안녕하세요, 존경하는 작가님, 저는 김나정을 정말 좋아합니다. Lora를 훈련시킬 수 있나요?

    gqmxmApr 20, 2023

    我帮你训练

    dudals364766May 6, 2023
    CivitAI

    respected author can you make astrid hofferson lora?

    AI_Anime_GirlsMay 21, 2023· 1 reaction
    CivitAI

    can we request son ye jin please :)

    1131May 22, 2023

    Yes!! Son Yejin would be perfect!

    musmancheema89May 24, 2023
    CivitAI

    hi Ithlinni,
    I am a free lance worker and I need a couple of lora's created for some project, I am willing to pay for those. Please let me know if you are interested and we can talk :)

    mostwantedkiss007Jun 4, 2023
    CivitAI

    Thank you! This IU lora works great!

    Unstable_NoobJun 4, 2023
    CivitAI

    Thank you for this beautiful lora model!

    llyzjk927Jun 8, 2023

    123

    Unstable_NoobJun 19, 2023
    CivitAI

    Can you please make one for Jung So-min. Please!

    oldroidSep 8, 2023· 1 reaction
    CivitAI

    How many images were used to train this model?

    SmoothBrainApeSep 13, 2023
    CivitAI

    We need and SDXL version of this! Love your work thanks for this!

    Unstable_NoobOct 14, 2023
    CivitAI

    Can you please make a SDXL version. Please!

    SmoothBrainApeDec 5, 2023· 2 reactions
    CivitAI

    9 months since this model was created, and still the best IU model out there. Still better than new SDXL models of IU. Thanks for making this model!

    jabsladerFeb 10, 2024
    CivitAI

    thank u for this and update plz

    Fathers_JobJan 29, 2025
    CivitAI

    인민여동생 ㅋㅋ

    shim6468970Feb 1, 2025

    ㅄㅋㅋ

    LORA
    SD 1.5
    by Ithlinni
    Download (Beta)
    girl
    photorealistic
    female
    asian
    woman
    celebrity
    korean
    realistic

    Details

    Downloads
    56,728
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    3/4/2023
    Updated
    5/7/2026
    Deleted
    5/23/2025
    Trigger Words:
    iu1

    Files

    iu_v35.safetensors

    Size:
    288.11 MB
    SHA256:
    c9598ba3470c9529ba2bb386a0bf74bcdcac5e6c4efcfde70067755f3b76316a

    Mirrors

    HuggingFace (58 mirrors)
    CivitAI (1 mirrors)
    TensorFiles (1 mirrors)
    CivitasBay (1 mirrors)

    Available On (1 platform)

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

    CivitasBay
    IU - IU V3.5