Trained on Tsubaki, 18 or 19 images.
LoRA on the new DiT V1. I'm quite happy that I don't have to use many LoRAs to achieve good results. Of course, it can always be better, but it's better than the SDXL, in my opinion
I only used Tsubaki to generate the images, but I think you can use other DiT models.
I use Euler A. 28 steps. Because the results are pretty good, I think you can generate 3 images in parallel, and you will still be fine. Unlike SDXL, DiT is pretty consistent from my initial experience. Of course, if you demand a lot of detail, then maybe go cheap first.
I use 3, 4.5 and 6CFG. It's a change from the previous SDXL and SD models.
LoRAs I used in generation: https://pixai.art/en/model/1839426804048327588
Only 1 "alien" LoRA so far. That's actually quite impressive, though, if some LoRA for better landscape details drop, then I might start using it too, to get an even better environment.
Thumbnail model: Tsubaki
