CivArchive
    Kwon Nara Disown Flux - v2.0
    Preview undefined

    Support Diswon on Ko-fi❤️ - Ko-fi ❤️ Where creators get support from fans through donations, memberships, shop sales and more! The original 'Buy Me a Coffee' Page.

    Join KPOP Diffusion Discord for learning/chatting: https://discord.gg/NrBDMX79Da

    Description

    FAQ

    Comments (6)

    Mr_JingujiDec 7, 2024
    CivitAI

    Oh mate, thx for version 2, Ima go and test it, I'll let you know the results.

    Mr_JingujiDec 7, 2024
    CivitAI

    Okay, this version is more stable than V1, but it still has its fair share of bad body posture. Out of 300 pics, I got 187 good ones. That's around 60% good results, but for some reason, your Seo Yeji Lora has a 95% success rate. It even improves the base Flux model, I'm obsessed with it.

    Disown8290
    Author
    Dec 8, 2024

    Yea this data set i used more full body images for better body posture but there isnt that many pics of her out there that are good quaily. I think this will be the best I can get for now

    Mr_JingujiDec 8, 2024

    @Disown8290 Yeah, that's probably the case, but how many pics are needed for a good result?

    Is it possible to get some of the perfect results from this Lora and add them to the previous dataset for training?

    Disown8290
    Author
    Dec 9, 2024

    @Mr_Jinguji I always use 30 pics so I can narrow down what I want. I only use real photos for dataset I haven't tired a outpit only yet. I don't think adding more photos make the dataset more flexible

    Mr_JingujiDec 10, 2024

    @Disown8290 I hope she does more HQ photoshoots so we can have a better dataset.

    LORA
    Flux.1 D

    Details

    Downloads
    120
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    12/6/2024
    Updated
    5/12/2026
    Deleted
    5/23/2025

    Files

    Kwon_Nara_Flux_Lora_v2.0.safetensors

    Mirrors