CivArchive
    Shadow Zelda | Goth Zelda (The Legend Of Zelda: Tears of the Kingdom) [Minimax H3] - v1.0
    NSFW

    Follow me on X/Twitter for exclusive content: https://x.com/Citron_Legacy 😁

    🎃 Happy Halloween! 👻

    I'm making Halloween Loras every day in October please follow my collection so you don't miss any treats! https://civarchive.com/collections/11841893

    Description

    zzGthZeldaC1tr0n, red eyes, short hair, black hair, pointy ears, crown braid,

    bare shoulders, black dress, hairclip, hair ornament, choker, vampire, pelvic curtain, sideboob, black thighhighs,

    <lora:GothZelda_H3_v1:1.0>,

    FAQ

    Comments (4)

    leonqubick591Oct 9, 2026
    CivitAI

    i dont get these character loras for h3?
    just...use a reference image. no need for a whole lora. unless you bake in voices and outfit types...which again i can just get through ai.

    CitronLegacy
    Author
    Oct 9, 2026

    Thanks for the comment!
    Trained motion and voices are definitely the most useful purpose of a H3 lora.
    Image only H3 loras have limited purpose, but with a lora like this
    1. You don't have to find different images for each image2video generation. You can get any angle/pose with just the lora.
    2. Its trained on the images I used in my dataset so that influences the art style and aspects of appearance (like body shape and facial expressions).

    3. It helps with different poses/expressions and consistent appearance at different angles.
    So for example my Natagoro lora can make different faces like this https://civitai.red/images/144708750

    4. From a usability/accessibility perspective this is easier for some users.

    leonqubick591Oct 12, 2026

    @CitronLegacy i just have GPT make a character sheet, front side back, then close up of face front and side, then zoomed in to eyes and mouth and accessories.
    then H3 just does its frikkin magic. over 6 clips tied together and its 99% the same character. no need for loras. its like....too easy nowadays so i think authors of these are used to the nightmare and headaches from before. i jsut have a whole library of characters made to pick from and it works so stupidly well. even with different styles. i did a anime style Aqua and used a real life ai photo of a hotel room and she interacted perfectly with no style melt in either separation. again, i think users are too used to the garbage we had to deal with soooo long ago (6 months)

    bvgaOct 29, 2026

    @leonqubick591 Even with multiple references MMH3 can have difficulty maintaining style and character features, especially if you have something particular. For instance, even with 6 references I'll occasionally get the character's skin becoming glossy, ribs being more bony than they should, etc, when cutting between shots. LoRAs like these can help keep things more consistent.

    LORA
    MiniMax H3

    Details

    Downloads
    190
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/7/2026
    Updated
    11/21/2026
    Deleted
    -
    Trigger Words:
    zzGthZeldaC1tr0n, red eyes, short hair, black hair, pointy ears, crown braid
    zzGthZeldaC1tr0n, red eyes, short hair, black hair, pointy ears, crown braid, bare shoulders, black dress, hairclip, hair ornament, choker, vampire, pelvic curtain, sideboob, black thighhighs

    Files

    GothZelda_H3_v1.safetensors

    Mirrors

    CivitAI (1 mirrors)