CivArchive
    Paste Crime - Mixed Media Graphic Style [KREA2] - v1.0
    NSFW
    Preview 143141314
    Preview 143141075
    Preview 143141085
    Preview 143141079
    Preview 143141084
    Preview 143141086
    Preview 143141087
    Preview 143141104
    Preview 143141090
    Preview 143141315
    Preview 143141089
    Preview 143141097
    Preview 143141093
    Preview 143141095
    Preview 143141102
    Preview 143141313
    Preview 143148064

    V2 NOTES:
    Paste Crime has had a fairly substantial rebuild under the hood. The original model worked, but after more testing with Krea2 I changed how I approach captioning, activation, prompting, and workflow support.

    CHANGES:

    Re-captioned AND re-trained from scratch.

    New natural-language activation phrase: A mixed-media editorial composition

    Removed the old arbitrary p@stecrime trigger from the revised training approach. No more nonsense tokens (I'm learning as I go!)

    Updated prompting guide. The new guide reflects how Paste Crime actually behaves after testing: photography, illustration, manga, collage, abstraction, typography, and editorial composition can all be used without forcing every image into the same recipe.

    Less “kitchen sink” prompting. Paste Crime works best when you give it a subject, a medium, and a few useful controls, then let the LoRA decide how to intervene.

    Wildcard support. I built a Paste Crime wildcard library from the model's visual vocabulary, including direct categories and higher-order presets for controlled randomization and faster exploration. Wildcards are available as an optional download. Just plug them into your custom wildcard folder and use a compatible wildcard workflow.
    ------------------------------------------------------------------------------------------------------
    Paste Crime doesn’t care what you start with, it just wants to interfere with it!!

    Built from a wildly varied mixed media dataset (including some of my own graphic design portfolio), Paste Crime can transform photography, manga, illustration, portraits, architecture, still life, animals, and whatever other nonsense you decide to throw at it.

    Rather than forcing everything into one fixed aesthetic, it reacts to the image itself. Depending on the subject and prompt, it can introduce hand drawn marks, ink and manga linework, pasted paper elements, crude doodles, graphic symbols, halftones, typography, bold color fields, photographic fragments, print textures, or full on collage chaos.

    That flexibility is the point.

    The A/B examples below use the same seeds and prompts. They show what Paste Crime is actually doing beneath the obvious collage elements: altering line quality, composition, graphic weight, texture, color, and the overall visual language of the image.

    Use it lightly. Push it hard. Give it something weird. Don't be afraid to try using it with other combinations of LoRA's. Give it hell.

    Included is a plug and play LLM prompting guide. Load it up into your preferred LLM, explain your subject or vision and let the prompt guide construct the perfect layout!
    ------------------------------------------------------------------------------------------------------
    Part of my experimental branch of graphic design layouts I'm trying to create in Krea2. While Krea can absolutely do editorial layouts, I'm creating these styles as specific types of graphical languages. New outfits for your designer to wear, so to speak.

    Check out the following A/B results to see how it performs:



    Description

    FAQ

    Comments (2)

    332orckSep 18, 2026· 1 reaction
    CivitAI

    Yoink, gotta mess around with this later

    LolTurdFerguson
    Author
    Sep 18, 2026

    MESS IT UP, DO TERRIBLE AND GREAT THINGS <3

    LORA
    Krea 2

    Details

    Downloads
    239
    Platform
    CivitAI
    Platform Status
    Available
    Created
    9/18/2026
    Updated
    9/28/2026
    Deleted
    -

    Files

    Paste_Crime_LLM_Prompting_Guide.zip

    Mirrors

    PasteCrimeKrea2.safetensors

    Mirrors

    HuggingFace (1 mirrors)
    CivitAI (1 mirrors)