Honestly, I don't fully understand what LoRa actually does. It was originally intended to be graffiti-style with flat images in a street art or sketch style. But I didn't achieve the desired result. What I ended up with, in my opinion, is a pretty good stylistic lore that adds a touch of sketch and shading. Anyway, check out the examples. Personally, I use this LoRa as an additional feature on small scales. Maybe someone will find it useful and interesting to play around with and experiment with. If there's interest, I'll try to expand and optimize the dataset for training in the future, when I get around to it.

Well, I usually use the following settings with this LoRa:

1. Euler A or DPM++ 2M SDE

2. Weight: 0.4-0.8

3. Resolution: 832x1216, 768x1344 (but it's optional, I guess)