CivArchive
    The I WonderMix - V3
    NSFW
    Preview 65552120Preview 65552228Preview 65552230Preview 65552430Preview 65552432Preview 65552477Preview 65552436Preview 65552221Preview 65552475Preview 65552224Preview 65552435Preview 65552439Preview 65622420Preview 66023621Preview 66232232

    I For Illustrious.

    This model is an Illustrious based merge, so the Fair AI Public License 1.0-SD is in use. Model used in the merge are linked to the version details. All credit goes to heir creators (thanks guy o/).

    Like all the Illustrious models, this is an anime model SFW/NSFW capable, with a ton of pro : excellent prompt adherence, very stable anatomy, a ton of concept, characters, artists and style available without LoRa. But also a bit of con depending of your prompt and scene : lack of details (especially Background), flat lighting and shaders.

    But overall it's a very nice model capable of a ton of things.


    Generation Settings

    All the images were generated with Forge UI. Only Text2Image with HiresFix were used.

    Euler a @ 30 Steps for base generation but, it's not necessarily the best, it's just the one that i liked the output the most. Euler and DPM++ 2M also have nice output, just with different level of details and a different color scheme.

    HiresFix :

    • 10 Steps

    • Denoising between 0.3 and 0.5 (0.4 being the nice spot)

    • Upscaler : None.... yup None. It's an underrated and under the radar option that deliver very nice output. Try it! You risk to be surprised ^^


    Prompt Settings

    This model need you to work on your prompt. While you can obtain nice result with a short prompt, the longer and more detailed your prompt is, the more your image will have details. if you want a list of all the tags go here.

    Don't be afraid to go over 75 tokens, and even 150 with this model, it won't break your output ^^

    Postivie Prompt

    Use the following at the start of your prompt

    (masterpiece, best quality, high quality, good quality)

    I tried a ton of things : at the start, at the end, with and without emphasis etc... the best output quality wise is the above.

    For the rest of the prompt, follow the structure below

    1girl/1boy/1other (or how many you want, or a mix of thoses three), Character/Serie/Artist, your prompt, style tags (3D, realistic, anime coloring, pastel color, neon palette, monochrome etc...).

    Negative Prompt

    (bad quality, low quality, worst quality:1.4), bad anatomy, bad hands, bad feet

    This is just the base negative prompt. you can add/remove others tags. Play with it to find the right one for the right generation. But this base work already very well in most case.


    Hope you will have fun and enjoy this model like i do!

    Don't hesitate to share your generation, i'm eager to see what others can get out of it!

    Description

    V3

    Model used in the merge :

    Baked VAE : XL_VAE_C

    Overall improvement. The style is a bit stronger than before.

    Use of Forge SAG with default settings on every images.

    It make thing more sharper and clearer.

    Checkpoint
    Illustrious

    Details

    Downloads
    481
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/24/2025
    Updated
    9/27/2025
    Deleted
    -