• check the images for prompts
• use lora at 0.5-4

using Pastel colors as tags also helps

Update : XL version is based on pony diffusion. havent tested on animagine. you guys can test and share feedback.**v3 is retrain of the cleaned dataset to get closer to original style

**v2 model is much clean. adjust the Lora weight if you need more style.

Use the preview images as a starting point

Lora weight of 1- works great

Highres fix is highly recommended.