Highly recommend using it on CivitAI's generator (1.3 - 1.6 strenght)
Base prompt: tsukidomo style illustration
Dataset: 500 images
Okay so this LoRa was trained on Flux Klein 4b, the biggest difference i have seen is the speed for the training dropping from 12 hours to a more friendly 6 hours time, which is a great time saver, anyways i'm still gonna play around with it to see where it lacks or completely breaks because believe me when i say it constantly creates artifacts and nonsensical elements to the image editing side.
Follow the original artist:
https://www.instagram.com/tsuukiyomo?utm_source=ig_web_button_share_sheet&igsh=ZDNlZDc0MzIxNw==
.jpeg)