Type | |
Stats | 145 26 167 |
Reviews | (22) |
Published | Sep 1, 2024 |
Base Model | |
Training | Steps: 4,000 |
Usage Tips | Strength: 1 |
Hash | AutoV2 548D001519 |
Disgaea style by Takehito Harada
This is a style LoRA for FLUX1.dev based on the artwork of Takehito Harada. He is best known for his work on the Disgaea video game series, but he also did original artwork and art for other video games like Labyrinth of Galleria/Refrain. The dataset is based on those three games and some original works and was extensivly tagged in natural language.
The LoRA knows no characters as they were not captioned (sorry, no easy Etna on this) its a pure style model.
Usage: To get the desired results prompt "anime style" or "digital illustration" with your prompt. Works well at strength 1. Without style prompts results will appear normal.
Hints: Try using additional style prompts like "oil painting" or "watercolor" as the LoRA mixes well with other styles. Works well with high guidance and SkimmedCFG/AutomatedCFG in high settings. Works well in a wide range of resolution, as it was trained with up to 1280x768 and 1024x1024 but also on lower resolutions.