After training different LORAs, this is the first model that can really fine detail of fine art styles.
This is absolutely crazy. The model can do so much more than just anime waifus or fake influencer photos.
These are all just single generations, no upscale, no inpainting, no refinement pass. Crazy!
Krea2 was labeled better than any other model with micro fashion details. You might have to look up different clothing styles, types of trims, necklines, fabric names, etc. but it actually gets the details.
The real power of that model is that it was labeled more accurately than any other model.
Qwen2512 still has better prompt adherence though for complex scenes like multiple characters.
But I can create fashions details with Krea2 that Qwen can't do because of the more accurate dataset captioning. Pretty cool!
China dominated with open source image models for so long;
Qwen, Z-Image, ERNIE, Wan, Hidream, Boogu
Now It seems American & Western AI companies are winning the AI image model race;
Anima, Krea2, Ideogram4, Flux Klein (German)
I am surprised to see American Ai open sourcing models that are better than the Chinese models. I always thought China's main motivation was undermine the USA market but now the West is open sourcing amazing models too.
LTX is actually an Israeli company by the way.
Can anyone score higher than 70 percent on this?
The model has some composition issues and crops images in weird place. The LoRAs don't quite as well as Qwen but DAMN! This model rocks. Insane levels of detail in single generations is possible.
Made my 1st Krea2 LoRA...
It is a style LoRA with a Dataset of over 600 images from Official Nikke Artwork.
It was NOT labeled with the characters name but will get the style.
You can download it here
https://civitai.com/models/2727496/nikke-goddess-of-victory-style-krea2?modelVersionId=3065893
This post contains content not supported on old Reddit. Click here to view the full post
can you tell if this is a real photo or Ai?
if you think it's real tell me why.
if you think it's ai tell me why.
Ideogram 4 Prompt Builder KJ node rocks.
you can make boxes on the canvas and 100 percent control the compotation.
Here is a link to the workflow. This workflow removes the censorship. It converts you text prompt into .json for you.
Best-friends theme in a peaceful park setting. Two women sitting close together on a bench, relaxed and happy, enjoying each other’s company. Soft sunshine, gentle breeze, warm and comforting mood.
Left woman
Hair & Makeup
Short, soft blonde hair with loose natural waves
Side bangs framing the face
Fresh, minimal makeup
Light blush and natural lip color
Soft, bright eyes with a warm smile
Attire
Short, light pastel dress with delicate lace trim
Hemline above the knee
Simple, elegant style with a relaxed fit
Matching light-colored heels with a delicate strap
Pose
Sitting slightly turned toward the right woman
Shoulder leaning gently toward her friend
Hands resting loosely on her lap
Body posture open and relaxed
Expression
Gentle, genuine smile
Warm, friendly, carefree mood
Eyes looking toward the viewer with kindness
Right woman
Hair & Makeup
Long, wavy red hair flowing over the shoulders
Soft side part with natural movement
Natural makeup with emphasis on fresh skin tone
Warm smile with a relaxed expression
Attire
Short, soft aqua or pastel dress with ribbon and lace accents
Hemline above the knee
Feminine, flowing style
light-colored heels complementing the dress
Pose
Sitting very close to the left woman
Head resting gently on her friend’s shoulder
One hand resting near her lap, relaxed posture
Legs crossed casually at the knee
Expression
Friendly, joyful smile
Content, connected feeling
Eyes looking toward the viewer with warmth
Background
Park setting with green grass and scattered trees
Sunlit atmosphere with dappled light through leaves
Wooden bench where they sit together
Soft-focus natural surroundings enhancing the peaceful mood
