This time besides checkpoint I've managed to make lora for you, hope you have fun! Sorry about it's size, but just the fact lora exists is already a big step forward for me.

My merges are more about style rather than prompt researching. Give it any prompt regarding who, where and what — and the render most of the time would be like previews. I'm downloading a lot of models here with great previews, but when I use my prompt they in best case give intermediate picture — nothing close to previews sometimes. Or trigger word is obligatory, which is not working for merges. Basic picture should be as given in previews if you pick the right base model. Omit style and author tags. You can add more lora's as you like (but not too much), to change it and have fun!

Recommended:

• CFG (3-5)

• 30 steps

• Face fix for small faces

• VAE backed (sdxl_vae)

Best regards!