>>938194152
>tips or know any guides

Know your checkpoint, not everything is equal, and will change LORAs. Keep to mainstream or well used checkpoints.

I use CivitAI to train my loras, you will need to have to have a few images as a dataset, collect as much as you can and trim bad examples (monochrome, WIPs, or sketches)

You will have to give each image you provide the booru tags associated with it. I use TAGGUI to automatically do it for me.
https://github.com/jhc13/taggui

Upload your images in a zip to CivitAI, select what you are trying to do, character, style, etc.
https://civitai.com/models/train

Train, use your default setting. They're pretty good. I reccomend taking 1 image booru tags and paste it as a prompt for the examples being made.

I wont go much farther but, in ComfyUI, I mix a few styles, using prompt weights to fine tune it.