CivArchive
    Despina the Pied Piper (Queen's Blade Grimoire) XL [Illustrious + Pony] - v1.0
    NSFW
    Preview 12687647
    Preview 12687646
    Preview 12687645
    Preview 12687638
    Preview 12687640
    Preview 12687644

    Despina, the Pied Piper, from Queen's Blade Grimore

    NOTABLE tags: purple hair, purple eyes, pointy ears, drill hair, hat, feather, Asymmetrical Clothes, revealing clothes, Vertical-Striped Clothes, Clothing Cutout, cape

    Description

    FAQ

    Comments (9)

    WorstAirtistMay 16, 2024
    CivitAI

    I saw you take requests, and so far I like your work, i find it thorough and quality compared to many. I was wondering if you'd be willing to do "Cus" (Kusu) from Dragonball Super, the short, petite 'angel.'

    there's 2 lora's for her atm and they both.... suck.

    neclordx
    Author
    May 16, 2024

    Sorry, I listen to suggestions about characters that are around my alley, and sometimes I do bounties that fancies my eye. But I do not take requests/comissions directly, unless then again is about characters I like already in first place.

    WorstAirtistMay 17, 2024

    :thumbsup:

    AndrachMay 25, 2024
    CivitAI

    May I know how many dataset images and steps to create this? Just for reference.

    neclordx
    Author
    May 25, 2024· 1 reaction
    Sure, is not my secret kung-fu technique or something. 35 images, 2 repeats, train batch size 1 for a total of 700 steps.
    AndrachMay 25, 2024

    @neclordx Thanks! People often recommend me to do 2k steps but it's sometimes ended up overcooked. Your confirmation helps me a lot!

    neclordx
    Author
    May 25, 2024

    @Andrach You need to consider though that all these QB loras count with really consistent datasets with the character in a great degree of poses and situations. Thats why they come generally so well even with low steps. Fore reference In other loras I try to not be under 40 images and the steps somewhere around 900-1000

    AndrachMay 26, 2024

    @neclordx Personally, what is your preferred minimum amount of dataset images when it comes to training Lora?

    neclordx
    Author
    May 26, 2024

    @Andrach 

    Basic loras: 40

    Multicostume: 30 from main costume + 20 of each additional costume

    Multiucharacter: 30 for character + 20 of them together

    Style: 90-100

    Now, this applies for PONY only. Which is a miracle. I've done pony loras for myself with about 20 pictures AI generated with a 1.5 lora and still got a nice XL pony lora.

    For standard XL (not pony) you may perfectly need 60-80 consistent images to get something reliable.

    LORA
    Pony

    Details

    Downloads
    475
    Platform
    CivitAI
    Platform Status
    Available
    Created
    5/16/2024
    Updated
    8/26/2026
    Deleted
    -

    Files

    Despina_QBG.safetensors

    Mirrors

    CivitAI (1 mirrors)

    Available On (1 platform)

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