Some time ago, when i was experimenting with my first LoRas of Stable Diffusion 1.5, accidentally, with a certain combination of models, I came across a new style whose results were very interesting to me. Specifically, it arose from the combination of my Alfonso Azpiri LoRa with the Lykon's LoRa 'More Details' and the RevAnimated checkpoint, and it was a semirealistic 2.5D style and a very flexible one, very good for fantasy and science-fiction. I have been thinking about training a LoRa of this style for a while on more modern models like Pony or Flux, so I have finally trained one for Flux from the best images generated by me with the aforementioned combination. I think the result is very good, also having the advantages of Flux such as better hands and better understanding of the prompts. As something curious, comment that it seems that the representation of the nipples has improved in this LoRa compared to the vanilla Flux.

Files with SHA256: 2e3b15e9f8c2ac2634509623285dce930988797c4714b79bf2a7645d59cf92f1