I made few mistakes with previous model (poor images in dataset mainly) that resulted in blurry outcomes. I retrained it properly this time and deleted the older versions. This is the proper 1.0 version of Simon Bisley style lora it creates beautiful vivid colorful illustration especially views are very pretty.Images are generated with euler+bong + refined with z-image turbo 8 steps on 0.3 no upscale.

Simon Bisley style

Its a test it seems to work tho i went full on ostrisai default for 2750 steps it preserves the medium well enough, the difference is pretty large when you compare lora and no lora gens. Minimax image is a crazy model 4k image in like a snap of a fingers really want to see if community adopt and develop tools around it.

bisley_000002750.safetensors

Files with SHA256: 908ab003ea516eec6c605d46d411bb44a7aed76b685e6151defd38b5d54f8b38