So apparently the LoRa that I trained on my own style works, I was just using it wrong because I fucked up during initial testing and then destroyed it with band-aid fixes.
Few basic quality prompts, a less overtrained epoch and medium weight makes it do exactly what it's supposed to. It's also recognizable even when mixed pretty strongly with other styles.
https://mega.nz/folder/F0JUTboT#ukdr0u-xPtR0JTVbTOn62g
Thilmer
2025-01-23 05:28:46 +0000 UTC