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.