A curated collection of unique art styles created by blending distinct artist tags within my Illustrious checkpoints. . It's just getting started but I'm hoping that this page will act as an experimental playground of diverse aesthetics and art styles.
Generation Settings
LoRA Strength: 0.8 – 1.0 (Recommended for the intended look)
Trigger Words: Not strictly required, but highly recommended if your prompt is struggling to pull the style through.
What’s Inside
Each version uploaded here represents a different stylistic experiment.
Flatso wasn't trained with prompts so it doesn't have an official trigger word but I find stylized illustration works well.
yuuCase Krea v1.1 uses the same dataset as the original but is trained without prompts. This greatly improved the models' prompt adherence but it means there's no trigger word (though digital illustration or just illustration work well as one).
Description
FAQ
Comments (6)
I love this lora. I've been looking for a style like this for a long time. However, it's a pretty big problem that the model's prompt tracking deteriorates radically above 0.5 strength. I've tried with several Krea2 checkpoints. Without lora, the prompt tracking is perfect, with lora it's only about. Has anyone experienced the problem?
Yeah, I've noticed that as well. i was thinking about trying another version without captioning the images to see if that helps. It seems to follow prompts better when I use the bypass lora
Was actually thinking about sending you a message about that, its only the yuuCase style I think
@KuroKenzo Yes, this only affects the yuuCase style. It would be nice if you could improve the prompt tracking, because I think it's the best lora in this style.
@fuzyz83591 @KuroKenzo I'll start training a new version today, hopefully I can get this fixed 🤞
@fuzyz83591 have you had a chance to try the new version?