Artsy Illustrious
All of the models I've created and published here, including this one, use a large, carefully curated dataset. I've repurposed it for highly stylized and detailed anime outputs and created a series of LoRA's in order to merge specific layers from each into Illustrious 2.0.
The result of this process was a substantial improvement of coherent subjects, backgrounds, and compositions, while maintaining the ever important illustrious prompt and style adherence.
BEST SETTINGS:
CFG: 5-6.5
STEPS: 30 - 50
SAMPLER: Euler Ancestral
SCHEDULER: SGM_UNIFORM
RESOLUTION: 1.25MP