This lora was trained on about 160 images from the game Granblue Fantasy. I wanted to see how this style would look like trained under the Pony Model and I also use this lora to try out some different training settings.
The new settings managed to create a stronger likeness to the source material. This lora likes to create elaborate backgrounds unprovoked. Sadly, this lora does an awful job at managing hands. Out of all the loras settings I tested, this was the only one that had convincing style and decent hands. (up close) I hope I can continue to improve my lora creation.
To get the results you see in the images above, these are the settings I used.
Prompt: score_9, score_8_up, score_7_up ,score_6_up, source_anime, [your prompt] <lora:Granblue_Fantasy_Style_XLv4:1>
Negatives: mosaic censoring,bar censor,censored,source_cartoon,source_pony,source_furry,simple background,
Sample: DPM++ 2S a
Steps: 33 (Whatever gives you the best results)
CFG Scale: 7
Highres Fix: 4xfoolhardyRemacri
Extension: [sd-webui-color-enhance] set to 0.5 (this increases the saturation a bit)
Please note, this is a style lora. It's not mean to recreate any of the characters from the game. Pony is bad at characters as well.
Even though this was trained on Pony, I exclusively use AutismMix and I recommend you do the same. All of the examples were made with that model.