Don't expect to get good results and maybe you'll be pleasantly surprised. I tossed a bunch of Star Wars renders, Dreamboothed them, and extracted to LoRA. The samples are mostly IMG2IMG using the RealisticVision 2 model and 100+ random poster size/style image generations I've made over the past 6 months or so.
If you use for TXT2IMG expect hit or miss results. For IMG2IMG I like denoise of ~.6 and simple prompt of "SWThings Artstyle" and few other terms. The samples came out of an overnight batch using -just- swthings artstyle in IMG2IMG applied against a folder of poster type generations.
Please don't dog me if this sucks for you - just find something else -wasn't going to dump it but works well w/ my workflow so for kicks.