A LoRA for a Darkwood-like style
All sample images are generated with my custom merged model UmbraMix
Positive prompts can be made simple for better abstraction and artistic freedom
Negative prompts should be as simple as possible such as (worst quality, low quality:1.4), (blurry:1.4).
All sample prompts are generate with a CLIP SKIP value of 2
I recommend using the fallowing weights:
aggressive: 0.8 ~ 0.9
soft / permissive: 0.35 ~ 0.55
I recommend using the kl-f8-anime2 VAE