The LoRa adds the distinct style of Zdzisław Beksiński to generated images.
Seems to be working best at the strength 0.7-1.5
LoRa is tested at:
20, 25, 35, 40, 50 and 60 steps
Guidance 2-15
My testing so far shows that the LoRa works well (for Flux .1 Dev) at relatively low guidance (~2) and 25-35 steps with LoRa strength 0.8-1.1
At this point the LoRa is not tested at Flux .1 Schnell or any of the fp8, nf4 and GGUF models.
I will update the description as more testing is done.
1. I diritti dei modelli ripubblicati appartengono ai creatori originali.
2. I creatori originali che desiderano reclamare il proprio modello devono contattare il personale di SeaArt AI tramite canali ufficiali. Clicca per rivendicare
