CivArchive
    Demon's Stele & Dog Princess - HappyLambBarn - Anima_V1_E7A1n2
    Preview 139509408
    Preview 139509496

    Introduction:

    Aug 20, 2025

    What could be more exciting than having another dog-girl character that I chose today?

    This princess—in a hentai-style game—features in a narrative about a hero who embarks on a journey to rescue her after she was trapped within a stone stele by a demon because of her own mistakes.

    I discovered this game online, downloaded it, and played while capturing screenshots to build a dataset. During the dataset creation, I modded the game’s textures to remove the background. This made it easier for AI training by reducing visual noise. Although I also have an Alpha 2 version, I’ll focus first on the original (version 1.07 Dev, unlocked all). The dataset images will focus solely on in-game visuals—not the introductory storytelling segments.

    Basically Alpha 2 just change styleart and implement more outfits nothing too much

    Since that day (August 20, 2025), I don't know why A1111 or ComfyUI on Colab has been constantly having errors and cannot be accessed, causing me to not be able to check which ecphos have sweet spots (aka learned best to avoid choosing insufficient quality) and create images before uploading. So this lora post will update the banner/preview image later.

    The tool I used for texture modding is "JPEXS Free Flash Decompiler" by jindrapetrik github


    Aug 12, 2026

    This is a version I ported to Anima to help flesh out this architectural ecosystem.

    In the "Anima v1" version, there are some issues regarding the accuracy of the outfit after generation. Here are the tags that do not match the original image:

    1. "armlet" often appears flat.

    2. "choker" turns out generic.

    3. "black straps shoulder" also turns out generic.

    Overall, the character accuracy is around 89%. Regarding the training process, the dataset included the "Alpha 2 Unity" version (which features softer lines compared to "Alpha 1 Flash"); this excludes story-telling cutscenes (as they were not part of the plan). I have no further comments.


    Prompt Trigger:

    Base Character:

    h1me, 1girl, solo, blonde hair, twintail hair, long twintall, green eyes, blush, hair between eyes, red bow on twintail, mini crown, tilted headwear, small breasts

    Base Character (Anima):

    h1me, 1girl, solo, dog girl, animal ears, dog ears, blonde hair, twintail hair, long twintall, green eyes, blush, hair between eyes, dog tail, red bow, bow on twintail, simple crown king, tilted headwear, thick eyebrows, short eyebrows,

    Outfit Sleeves:

    outfit_casual, dress, red dress, long sleeves, off shoulder, black straps shoulder, frilled skirt, layered skirt, black layered skirt, ribbon, armlet, white thighhigh, scallop-top, scallop-top thigh-high socks, choker

    Outfit Sleeves (Anima):

    outfit_casual, gold choker, dress, red dress, long sleeves, puffy sleeves, off shoulder, gold buttons, black straps shoulder, frilled skirt, layered skirt, black skirt, ribbon, armlet, white thighhigh, scallop-top, scallop-top thigh-high

    Outfit Sleevesless:

    outfit_casual_sleeveless, dress, red dress, sleeveless dress, sleeveless, frilled dress, frills, collarbone, bare shoulders, bare arms, white thighhigh, scallop-top, scallop-top thigh-high socks, layered skirt, choker, black layered skirt, black straps shoulder

    Outfit Sleevesless (Anima):

    outfit_casual_sleeveless, gold choker, dress, red dress, sleeveless, off shoulder, gold buttons, black straps shoulder, frilled skirt, layered skirt, black skirt, ribbon, white thighhigh, scallop-top, scallop-top thigh-high

    Description

    ### Training Information (Anima DiT)

    + Output Name: bf16_Anima_V1_Hime

    + Network Module: networks.lora_anima

    + Epochs: 7

    + Max Train Steps: Auto (Epoch)

    + Learning Rate: 0.0001

    + Text Encoder LR: 5e-05

    + Optimizer: AdamW8bit

    + Mixed Precision: bf16

    + LR Scheduler: constant

    + LR Warmup Steps: 0

    + Rank: 16/16 (Dim, Alpha)

    ### Anima Arguments:

    + timestep_sample_method: logit_normal

    + discrete_flow_shift: 3

    + weighting_scheme: logit_normal

    ### Dataset Info (dataset.toml):

    + Batch Size: 2

    + Resolution: 1024x1024

    + Steps (calculated): 210

    + Bucketing: True

    ### Datasets Parameter:

    + 4_gernal : images: 20 | repeats: 3 | keep_tokens: 1 | flip_aug: False | shuffle_caption: True

    FAQ

    LORA
    Anima

    Details

    Downloads
    10
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/12/2026
    Updated
    8/12/2026
    Deleted
    -
    Trigger Words:
    h1me, 1girl, solo, dog girl, animal ears, dog ears, blonde hair, twintail hair, long twintall, green eyes, blush, hair between eyes, dog tail, red bow, bow on twintail, simple crown king, tilted headwear, thick eyebrows, short eyebrows,
    outfit_casual, gold choker, dress, red dress, long sleeves, puffy sleeves, off shoulder, gold buttons, black straps shoulder, frilled skirt, layered skirt, black skirt, ribbon, armlet, white thighhigh, scallop-top, scallop-top thigh-high
    gold choker, dress, red dress, sleeveless, off shoulder, gold buttons, black straps shoulder, frilled skirt, layered skirt, black skirt, ribbon, white thighhigh, scallop-top, scallop-top thigh-high
    h1me

    Files

    bf16_Anima_V1_Hime.safetensors

    Mirrors