KREA trained their own AI ART models from scratch -- and just released "
KREA 2 Large" and "
KREA 2 Medium" on May 12th, 2026 (
for subscribers on the more expensive plans) -- for a more whimsical
aesthetic coherence. This week however they let all paid subscribers use it for FREE for one week (
May 18 - 24, 2026).
The above video was created with AI in:
"Only this week:
UNLIMITED Krea 2 generations for all paid plans."
(May 18 - 24, 2026)
M O O D B O A R D S
The key is to upload a bunch of images to make "Moodboards," and then generate images from
TEXT prompts.
Sometimes uploading and additional reference image helps:Moodboard made from 24 images of my Paper Figure sculptures
-- which I made public
My Paper Figure Sculptures
described in words
GENERATIONS
"MONOCHROMATIC BIOMECHANICAL STATUARY"
I made another sculptural Moodboard -- "
MONOCHROMATIC BIOMECHANICAL STATUARY" -- with 10 images that were
based on Bernini sculptures and my drawings, which I started generating in
Leonardo.AI back in 2023. However the results were not always that exciting:
MONOCHROMATIC BIOMECHANICAL STATUARY
Moodboard based on 10 of my drawings
in the style of Bernini
GENERATIONS
(just OK results)
I cannot get
Midjourney or the new "
KREA 2" models (
designed to compete with Midjourney) to generate anything decent in my KRRRL drawing style. The premiere AI ART programs do not work well generating organic forms with lines; perhaps they even censor those forms.
I uploaded my figure drawing and used the new "KREA 2 Large" model to alter my drawing, but did not like the results:
My drawing (left)
and that drawing altered by the new "KREA 2 Large" model (right)
M O O D B O A R D
FAILS
Nor did I like the results using the Moodboards made from my figure drawings:
"FRAGMENTED GESTURAL MINIMALISM"
I made a Moodboard based on 11 of my more recent raw figure drawings --
"FRAGMENTED GESTURAL MINIMALISM" -- but I did not like the generated results:
"FRAGMENTED GESTURAL MINIMALISM"
based on 11 of my more recent figure drawings
I did not like the results
from the Moodboard based on my raw drawings
"DECONSTRUCTED GEOMETRIC FUTURISM"
I made a Moodboard based on 10 of my drawings that have been converted into "
black and white dotted line" drawings --
"DECONSTRUCTED GEOMETRIC FUTURISM" -- but I did not like the generated results:
DECONSTRUCTED GEOMETRIC FUTURISM
based on 10 of my "dotted line" drawings
I did not like the results from the Moodboard
made from my slightly altered figure drawings,
in a "dotted line" style
Nor did I like the results from my FAVORITE MOODBOARD --
"Fragmented Structural Papercraft" -- when I uploaded one of my figure drawings as a reference:
"Fragmented Structural Papercraft"
I uploaded the first drawing in this blog
and altered it with the "KREA2 AI" models
using my "Fragmented Structural Papercraft" Moodboard
The results did not look like my drawing or the images in my Moodboard -- not when using either the "K2 Large" or the "K2 Medium" AI models:
"K2 Large" model result
KREA asked me for feedback on May 20th, and I told them that the new models did not work well for line drawings, especially with organic forms. However I did say that I liked what the Moodboards were doing otherwise, and thanked them for a free week to experiment.
Likewise Google never gives me any results near my drawing style when I do a reverse search with Google Lens:
The "search" result drawings on the left
do not much resemble
my reference drawing on the right
HOWEVER Google did do a decent job describing my approach to making a figure drawing:
HOW to PRACTICE THIS STYLE
1. Start with Gestures: Capture the flow and tilt of the spine and hips using simple stick lines first
2. Box the Torso: Treat the chest and pelvis as two solid bricks or wedges that can twist independently
3. Carve the Planes: Instead of drawing smooth outer outlines, connect points using straight, deliberate lines to "carve" out the muscles
4. Wrap the Surface: Draw light horizontal and vertical bands around the arms and legs like bandages to force yourself to see their thickness
- I might add that I like to drag lines through the image to "translate the force" that act on the figure.
*****
Google upgraded their new video generator to "Omni" (for paid users), and announced new rules when I opened
FLOW on May 19th, 2026:
.gif)