CivArchive
    BlueFantasy - v1.0
    NSFW
    Preview 144314654
    Preview 144314653
    Preview 144314657
    Preview 144314659
    Preview 144314663
    Preview 144314666
    Preview 144314668
    Preview 144314671
    Preview 144314674
    Preview 144314675
    Preview 144314677
    Preview 144314679
    Preview 144314684
    Preview 144314687
    Preview 144314690
    Preview 144314694
    Preview 144314697
    Preview 144314701

    This lora is trained on fantasy artwork which has a cool tone, and a traditional fantasy feel. No special prompting but you must put "illustration" or "painting" or you will get photos! See examples for style,

    Description

    FAQ

    Comments (6)

    MirabilisOct 2, 2026
    CivitAI

    looks cool, but no prompt examples?

    robadams171
    Author
    Oct 9, 2026

    My prompts rarely get beyond 30 words, with prompting I feel less is more. You can use Krea influence images using the the Text Encode Qwen Image Edit node, they tend to work better for conditioning than a long prompt which mentions "etherial" a lot!

    robadams171
    Author
    Oct 9, 2026· 1 reaction

    I've added an image with a prompt to show the structure that Krea likes. Remember every word you add weakens the influence of the other words. You do need a certain amount of description to get good quality, but the long LLM prompts are not great IMO.

    MirabilisOct 9, 2026

    @robadams171 Cool. I'm testing some stuff at present, but I'll give your lora some attention over the weekend. I generally prompt long (200-500+ tokens) but then I tend to do a lot of photography, but my recent dying earth set was Fantasy, and I have some other fantasy sets in mind.

    robadams171
    Author
    Oct 10, 2026· 1 reaction

    looked at your page some very nice images! I have a couple of fashion loras I'll post later on. One on 60's fashion the other Polaroid grunge. I do illustration and I find long prompts weaken the lora style, especially if it's a unique one that the base model doesn't know.

    MirabilisOct 10, 2026

    @robadams171 Thank you. For the photography, I'm playing a bit with Loras, but for the most part I'm using a pretty robust prompting system I developed over the last few months using a ChatGPT project. Rather than using visual styles, it uses distilled compositional grammars (a lot of research files), using everything from artists, photographers, magazines, coupled with fictive models/Actresses. It gets away from the normal AI framing, especially if you use artists I've found.

    Then we come up with an idea (for instance, the latest is Victoria in a Vogue-style shoot in a museum), and just iterate ion the outputs until I'm happy with the results.

    I'm burning through some old generational sets at the present, which I uploaded a week or so ago. However, I'm going be uploading some newer ones soon, which are much more detailed and material-rich.

    However going to do some fantasy stuff as well.

    LORA
    Krea 2

    Details

    Downloads
    121
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/30/2026
    Updated
    11/9/2026
    Deleted
    -

    Files

    bluefantasy_v1.safetensors

    Mirrors

    CivitAI (1 mirrors)