CivArchive
    Chaos Style - Illustrious v1
    Preview 51311915Preview 51311916Preview 51311914Preview 51311903Preview 51311942Preview 51311898Preview 51311902Preview 51311907Preview 51311904Preview 51311897Preview 51311920Preview 51311911Preview 51311908Preview 51311921Preview 51311912Preview 51311918Preview 51311917Preview 51311906Preview 51311905Preview 51311919

    I am unsure and unaware of any artist who uses a style similar to this. I am under the impression that this is my own unique and original style. There was no artist or style reference while making this lora other than my own image's from my own checkpoints.

    Illustrious V1
    Attempt at getting a similar style, look, and feel of the V1 and V2 versions onto Illustrious. Due to issues with training, I used a various mix of v1, v2, and gen 4 and gen 5 of my illustrious training's to create this version. Check version notes for the full recipe.

    V2
    Focus on quality, detail, and coherence. This version works well on humans and anthro creatures but struggles with "pony" characters. you may need to reduce weight for these characters.

    Detail and quality should be double of V1.

    V1

    Focus of style is larger eyes, detailed eyes, cute and slightly deformed body proportions, mystical, magical, and effects. This model does work on SDXL and Pony models. Tested with Anime, sudo realistic, and photo, realistic models.

    Prompting / Triggers

    Trigger word is "chaosstyle" but was designed to work without this keyword.

    Positive:

    Large detailed eyes, magic, magical effects, majestic, cute, adorable,

    Negative:

    score_4, score_5, score_6,
    watermark, logo, signature, writing, boring,
    (hands:1.5), ugly, low res,

    Description

    Trained 5 different lora sets at 150 lora total. Found v4 and v5 as the best.

    Used those in different configurations as defined below: Merge v1 and v2 Pony at 75% and 100% then merge v4 and v5 Illu which are custom trainings. Several epochs were used, feel free to check each version as defined "e##". These are trained at 60 steps each so e20 would be steps 1200. These items were merged using ComfyUI going by model strength / clip.

    ChaosStyle_Illu_M_v4_p100_e20 1.40 / 0.50
    ChaosStyle_Illu_M_v5_p75_e20 1.40 / 0.50
    -Dare_Linear 0.5 Density
    ChaosStyle_Illu_M_v5_p75_e21 1.50 / 0.50
    -Dare_Linear 0.5 Density
    --Add Density 1.0
    ---ChaosStyle_Illu_M_v1

    ChaosStyle_Illu_M_v4_p75_e21 0.75 / 0.90
    ChaosStyle_Illu_M_v4_p100_e20 0.75 / 0.90
    -Dare_Linear 0.75 Density
    ChaosStyle_Illu_M_v5_p75_e20 0.75 / 0.90
    ChaosStyle_Illu_M_v5_p75_e21 0.75 / 0.90
    --Dare_Linear 0.75 Density
    ---Add Density 1.0
    ----ChaosStyle_Illu_M_v2

    ChaosStyle_Illu_M_v1 1.20 / 0.90
    ChaosStyle_Illu_M_v2 0.40 / 0.90
    -Dare_Ties 0.6
    --ChaosStyle_Illu_M_v4

    ChaosStyle_Illu_M_v1 1.50 / 0.50
    ChaosStyle_Illu_M_v4 0.15 / 0.50
    -Dare_Ties 1.0
    --ChaosStyle_Illu_M_v5

    ChaosStyle_Illu_M_v5 Resized with Dare_Linear_SVD
    -Rank 32 / Rank SVD 32
    --ChaosStyle_Illu_M_v5_Resized5

    ChaosStyle_Illu_M_v5_Resized5 -> Renamed to ChaosStyle_Illu_v1

    LORA
    Illustrious

    Details

    Downloads
    95
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/13/2025
    Updated
    9/28/2025
    Deleted
    -
    Trigger Words:
    chaosstyle

    Files

    ChaosStyle_Illu_v1.safetensors

    Mirrors