Several subsets of a larger dataset I have curated for a later larger lora.
This covers photorealistic genre outputs. Specifically cyberpunk, fantasy, Victorian, post-apocalyptic and old west.
Using cinematic high quality high concept as descriptors or thereabouts will tend to steer towards higher quality outputs. You should be able to be very specific.
I set this to run up with v48 here, because it's been more stable on tensor art. I also used that for the sample pictures. BUT you are running locally use the proper final version.
A larger more intensive train using all this data, and more, will be forthcoming at some point. Although I have not settled on a base model for sure for that (I'll do two versions, one roughly flux based one SDXL based).
Enjoy :)